vu university amsterdam - the social web 2016 - lecture 5
TRANSCRIPT
Social Web2016
Lecture 5 Personalization on the Social Web
Lora Aroyo and Davide Ceolin(some slides adapted from Fabian Abel)
The Network InstituteVU University Amsterdam
theory amp techniques for how to design amp evaluate
recommenders amp user models to use in Social Web applications
Social Web 2016 Lora Aroyo and Davide Ceolin
Fig 1 Functional model of tasks and sub-tasks specifically suited for SASs
Functional model of tasks and sub-tasks specifically suited for SASs (Ilaria Torre 2009)
Social Web 2016 Lora Aroyo and Davide Ceolin
Kevin Kelly
How to infer amp represent user information that supports a given application or context
User Modeling
Social Web 2016 Lora Aroyo and Davide Ceolin
bull Application has to obtain understand amp exploit information about the user
bull Information (need amp context) about user
bull Inferring information about user amp representing it so that it can be consumed by the application
bull Data relevant for inferring information about user
User Modeling Challenge
Social Web 2016 Lora Aroyo and Davide Ceolin
bull People leave traces on the Web and on their computersbull Usage data eg query logs click-through-data bull Social data eg tags (micro-)blog posts comments
bookmarks friend connections bull Documents eg pictures videosbull Personal data eg affiliations locations bull Products applications services - bought used installed
bull Not only a userrsquos behavior but also interactions of other users bull ldquopeople can make statements about merdquobull ldquopeople who are similar to me can reveal information
about merdquobull ldquosocial learningrdquo collaborative recommender systems
User amp Usage Datais Everywhere
Social Web 2016 Lora Aroyo and Davide Ceolin
bull User Profile = data structure = a characterization of a user at a particular moment represents what from a given (system) perspective there is to know about a user The data in the profile can be explicitly given by user or derived by system
bull User Model = definitions amp rules for the interpretation of observations about the user amp about the translation of that interpretation into the characteristics in a user profile user model is the recipe for obtaining amp interpreting user profiles
bull User Modeling = the process of representing the user
UM Basic Concepts
Social Web 2016 Lora Aroyo and Davide Ceolin
bull Overlay User Modeling describe user characteristics eg ldquoknowledge of a userrdquo ldquointerests of a userrdquo with respect to ldquoidealrdquo characteristics
bull Customizing user explicitly provides amp adjusts elements of the user profile
bull User model elicitation ask amp observe the user learn amp improve user profile successively ldquointeractive user modelingrdquo
bull Stereotyping stereotypical characteristics to describe a user
bull User Relevance Modeling learninfer probabilities that a given item or concept is relevant for a user
Related scientific conference httpumap2011org Related journal httpumuaiorg
User Modeling Approaches
Social Web 2016 Lora Aroyo and Davide Ceolin
httpfarm7staticflickrcom62406346803873_e756dd9bae_bjpg
Which approach suits best the conditions of
applications
bull among the oldest user modelsbull used for modeling student
knowledgebull the user is typically
characterized in terms of domain concepts amp hypotheses of the userrsquos knowledge about these concepts in relation to an (ideal) expertrsquos knowledge
bull concept-value pairs
Overlay User Models
Social Web 2016 Lora Aroyo and Davide Ceolin
bull Ask the user explicitly learnbull NLP intelligent dialoguesbull Bayesian networks Hidden Markov models
bull Observe the user learn bull Logs machine learningbull Clustering classification data mining
bull Interactive user modeling mixture of direct inputs of a user observations and inferences
User Model Elicitation
Social Web 2016 Lora Aroyo and Davide Ceolin
httphunchcomSocial Web 2016 Lora Aroyo and Davide Ceolin
Social Web 2016 Lora Aroyo and Davide Ceolin
httpfarm1staticflickrcom155413650229_31ef379b0b_bjpg
bull set of characteristics (eg attribute-value pairs) that describe a group of users
bull user is not assigned to a single stereotype - user profile can feature characteristics of several different stereotypes
User Stereotypes
based on slides from Fabien Abel
based on slides from Fabien Abel
Can we infer a Twitter-based User
Profile
User Modeling Building Blocks
based on slides from Fabien Abel
User Modeling Building Blocks
based on slides from Fabien Abel
User Modeling Building Blocks
based on slides from Fabien Abel
User Modeling Building Blocks
based on slides from Fabien Abel
Observationsbull Profile characteristics
bull Semantic enrichment solves sparsity problemsbull Profiles change over time recent profiles reflect
better current user demandsbull Temporal patterns weekend profiles differ
significantly from weekday profilesbull Impact on recommendations
bull The more fine-grained the concepts the better the recommendation performance entity-based gt topic-based gt hashtag-based
bull Semantic enrichment improves recommendation quality
bull Time-sensitivity (adapting to trends) improves performance
Social Web 2016 Lora Aroyo and Davide Ceolin
User Modelingit is not about putting everything in a user profile
it is about making the right choices
Social Web 2016 Lora Aroyo and Davide Ceolin
User Adaptation
Knowing the user to adapt a system or interfaceto improve the system functionality and user
experience
Social Web 2016 Lora Aroyo and Davide Ceolin
A Jameson Adaptive interfaces and agents The HCI handbook fundamentals evolving technologies and emerging applications pp 305ndash330 2003
User-Adaptive Systems
Social Web 2016 Lora Aroyo and Davide Ceolin
based on slides from Fabien Abel
Lastfm adapts to your music taste
bull Overfitting ldquobubble effectsrdquo loss of serendipity problem bull systems may adapt too strongly to the
interestsbehaviorbull eg an adaptive radio station may always play the
same or very similar songsbull We search for the right balance between novelty and
relevance for the userbull ldquoLost in Hyperspacerdquo problem
bull when adapting the navigation ndash ie the links on which users can click to findaccess information
bull eg re-orderinghiding of menu items may lead to confusion
Issues in User-Adaptive Systems
Social Web 2016 Lora Aroyo and Davide Ceolin
What is good user modelling amp
personalisation
httpwwwflickrcomphotosbellarosebyliz4729613108
From the consumer perspective of an adaptive system
From the provider perspective of an adaptive system
Success Perspectives
Social Web 2016 Lora Aroyo and Davide Ceolin
bull User studies askobserve (selected) people whether you did a good job
bull Log analysis Analyze (click) data and infer whether you did a good job
bull Evaluation of user modelingbull measure quality of profiles directly eg measure
overlap with existing (true) profiles or let people judge the quality of the generated user profiles
bull measure quality of application that exploits the user profile eg apply user modeling strategies in a recommender system
Evaluation Strategies
Social Web 2016 Lora Aroyo and Davide Ceolin
Evaluating User Modeling in RecSys
Social Web 2016 Lora Aroyo and Davide Ceolin
Possible Metrics
Social Web 2016 Lora Aroyo and Davide Ceolin
bull The usual IR metricsbull Precision fraction of retrieved items that are relevantbull Recall fraction of relevant items that have been
retrievedbull F-Measure (harmonic) mean of precision and recall
bull Metrics for evaluating recommendation (rankings)bull Mean Reciprocal Rank (MRR) of first relevant itembull Successk probability that relevant item occurs
within the top kbull If a true ranking is given rank correlations bull Precisionk Recallk amp F-Measurek
bull Metrics for evaluating prediction of user preferencesbull MAE = Mean Absolute Errorbull TrueFalse PositivesNegatives
Predict relevantusefulinteresting items for a given user (in a given context) itrsquos often a ranking task
Recommendation Systems
Social Web 2016 Lora Aroyo and Davide Ceolin
March 28 2013
Social Web 2016 Lora Aroyo and Davide Ceolin
Social Web 2016 Lora Aroyo and Davide Ceolin
Social Web 2016 Lora Aroyo and Davide Ceolin
httpwwwwiredcommagazine201111mf_artsyall1
Social Web 2016 Lora Aroyo and Davide Ceolin
Collaborative Filteringbull Memory-based User-Item matrix ratingspreferences of users =gt
compute similarity between users amp recommend items of similar users
bull Model-based Item-Item matrix similarity (eg based on user ratings) between items =gt recommend items that are similar to the ones the user likes
bull Model-based Clustering cluster users according to their preferences =gt recommend items of users that belong to the same cluster
bull Model-based Bayesian networks P(u likes item B | u likes item A) = how likely is it that a user who likes item A will like item B learn probabilities from user ratingspreferences
bull Others rule-based other data mining techniques
Social Web 2016 Lora Aroyo and Davide Ceolin
bull complete input data is required
bull pre-computation not possible
bull does not scale well bull high quality of
recommendations
bull abstraction (model) of input data
bull pre-computation (partially) possible (model has to be re-built from time to time)
bull scales betterbull abstraction may reduce
recommendation quality
Memory vs Model-based
Social Web 2016 Lora Aroyo and Davide Ceolin
bull collaborative filtering lsquoneighborhoodsrsquo of people with similar interest amp recommending items based on likings in neighborhood
bull limitations next to lsquocold startrsquo and lsquosparsityrsquo the lack of control (over onersquos neighborhood) is also a problem ie cannot add lsquotrustedrsquo people nor exclude lsquostrangersquo ones
bull therefore interest in lsquosocial recommendersrsquo where presence of social connections defines the similarity in interests (eg social tagging CiteULike)bull does a social connection indicate user interest similaritybull how much users interest similarity depends on the strength
of their connectionbull is it feasible to use a social network as a personalized
recommendation[Lin amp Brusilovsky Social Networks and Interest Similarity The Case of CiteULike HTrsquo10]
Social Networks amp Interest Similarity
bull unilaterally connected pairs have more common itemsmetadatatags than non-connected pairsbull highest similarity for direct connections - decreasing with the increase of distance between users in SN bull reciprocal relationship users - significantly larger similarity than users in a unidirectional relationship bull traditional item-level similarity may be less reliable to find similar users in social bookmarking systemsbull peers connected by self-defined social connections could be a useful source for cross-recommendation
Conclusions
Social Web 2016 Lora Aroyo and Davide Ceolin
bull Input characteristics of items amp interests of a user into characteristics of items =gt Recommend items that feature characteristics which meet the userrsquos interests
bull Techniquesbull Data mining methods Cluster items based on their
characteristics =gt Infer usersrsquo interests into clustersbull IR methods Represent items amp users as term vectors
=gt Compute similarity between user profile vector and items
bull Utility-based methods Utility function that gets an item as input the parameters of the utility function are customized via preferences of a user
Content-based Recommendations
Social Web 2016 Lora Aroyo and Davide Ceolin
based on slides from Fabien Abel
Content Features
based on slides from Fabien Abel
User Model
based on slides from Fabien Abel
Recommendations
RecSys Issuesbull Cold-start problem (new user problem) nolittle data available to infer preferences of new
usersbull Changing User Preferences user interests may change over timebull Sparsity problem (new item problem) item descriptions are sparse eg not many user
rated or tagged an itembull Lack of Diversity (overfitting) when adapting too strongly to the preferences of users they
might see samesimilar recommendationsbull Use the right context users do things which might not be relevant for their user model
eg try out things do stuff for other peoplebull Research challenge right balance between serendipity amp personalizationbull Research challenge right way to use the influence of recommendations on userrsquos behavior
Social Web 2016 Lora Aroyo and Davide Ceolin
Announcementsbull Next deadlinesbull Tuesday March 1st 2359
Assignment 2bull Friday March 4th 1000 Post
questionbull Friday March 4th 1700 Vote
questions
image source httpwwwflickrcomphotosbionicteaching1375254387
Hands-on Teaser
bull Your Facebook Friendsrsquo popularity in a spread sheet
bull Locations of your Facebook Friendsbull Tag Cloud of your wall posts
- Social Web 2016
- Slide 2
- Slide 3
- Slide 4
- Slide 5
- Slide 6
- Slide 7
- Slide 8
- Slide 9
- Slide 10
- Slide 11
- Slide 12
- Slide 13
- Slide 14
- Slide 15
- Slide 16
- Slide 17
- Slide 18
- Slide 19
- Slide 20
- Slide 21
- Slide 22
- Slide 23
- Slide 24
- Slide 25
- Slide 26
- Slide 27
- Slide 28
- Slide 29
- Slide 30
- Slide 31
- Slide 32
- Slide 33
- Slide 34
- Slide 35
- Slide 36
- Slide 37
- Slide 38
- Slide 39
- Slide 40
- Slide 41
- Slide 42
- Slide 43
- Slide 44
- Slide 45
- Slide 46
- Slide 47
- Slide 48
- Announcements
- Slide 50
-
theory amp techniques for how to design amp evaluate
recommenders amp user models to use in Social Web applications
Social Web 2016 Lora Aroyo and Davide Ceolin
Fig 1 Functional model of tasks and sub-tasks specifically suited for SASs
Functional model of tasks and sub-tasks specifically suited for SASs (Ilaria Torre 2009)
Social Web 2016 Lora Aroyo and Davide Ceolin
Kevin Kelly
How to infer amp represent user information that supports a given application or context
User Modeling
Social Web 2016 Lora Aroyo and Davide Ceolin
bull Application has to obtain understand amp exploit information about the user
bull Information (need amp context) about user
bull Inferring information about user amp representing it so that it can be consumed by the application
bull Data relevant for inferring information about user
User Modeling Challenge
Social Web 2016 Lora Aroyo and Davide Ceolin
bull People leave traces on the Web and on their computersbull Usage data eg query logs click-through-data bull Social data eg tags (micro-)blog posts comments
bookmarks friend connections bull Documents eg pictures videosbull Personal data eg affiliations locations bull Products applications services - bought used installed
bull Not only a userrsquos behavior but also interactions of other users bull ldquopeople can make statements about merdquobull ldquopeople who are similar to me can reveal information
about merdquobull ldquosocial learningrdquo collaborative recommender systems
User amp Usage Datais Everywhere
Social Web 2016 Lora Aroyo and Davide Ceolin
bull User Profile = data structure = a characterization of a user at a particular moment represents what from a given (system) perspective there is to know about a user The data in the profile can be explicitly given by user or derived by system
bull User Model = definitions amp rules for the interpretation of observations about the user amp about the translation of that interpretation into the characteristics in a user profile user model is the recipe for obtaining amp interpreting user profiles
bull User Modeling = the process of representing the user
UM Basic Concepts
Social Web 2016 Lora Aroyo and Davide Ceolin
bull Overlay User Modeling describe user characteristics eg ldquoknowledge of a userrdquo ldquointerests of a userrdquo with respect to ldquoidealrdquo characteristics
bull Customizing user explicitly provides amp adjusts elements of the user profile
bull User model elicitation ask amp observe the user learn amp improve user profile successively ldquointeractive user modelingrdquo
bull Stereotyping stereotypical characteristics to describe a user
bull User Relevance Modeling learninfer probabilities that a given item or concept is relevant for a user
Related scientific conference httpumap2011org Related journal httpumuaiorg
User Modeling Approaches
Social Web 2016 Lora Aroyo and Davide Ceolin
httpfarm7staticflickrcom62406346803873_e756dd9bae_bjpg
Which approach suits best the conditions of
applications
bull among the oldest user modelsbull used for modeling student
knowledgebull the user is typically
characterized in terms of domain concepts amp hypotheses of the userrsquos knowledge about these concepts in relation to an (ideal) expertrsquos knowledge
bull concept-value pairs
Overlay User Models
Social Web 2016 Lora Aroyo and Davide Ceolin
bull Ask the user explicitly learnbull NLP intelligent dialoguesbull Bayesian networks Hidden Markov models
bull Observe the user learn bull Logs machine learningbull Clustering classification data mining
bull Interactive user modeling mixture of direct inputs of a user observations and inferences
User Model Elicitation
Social Web 2016 Lora Aroyo and Davide Ceolin
httphunchcomSocial Web 2016 Lora Aroyo and Davide Ceolin
Social Web 2016 Lora Aroyo and Davide Ceolin
httpfarm1staticflickrcom155413650229_31ef379b0b_bjpg
bull set of characteristics (eg attribute-value pairs) that describe a group of users
bull user is not assigned to a single stereotype - user profile can feature characteristics of several different stereotypes
User Stereotypes
based on slides from Fabien Abel
based on slides from Fabien Abel
Can we infer a Twitter-based User
Profile
User Modeling Building Blocks
based on slides from Fabien Abel
User Modeling Building Blocks
based on slides from Fabien Abel
User Modeling Building Blocks
based on slides from Fabien Abel
User Modeling Building Blocks
based on slides from Fabien Abel
Observationsbull Profile characteristics
bull Semantic enrichment solves sparsity problemsbull Profiles change over time recent profiles reflect
better current user demandsbull Temporal patterns weekend profiles differ
significantly from weekday profilesbull Impact on recommendations
bull The more fine-grained the concepts the better the recommendation performance entity-based gt topic-based gt hashtag-based
bull Semantic enrichment improves recommendation quality
bull Time-sensitivity (adapting to trends) improves performance
Social Web 2016 Lora Aroyo and Davide Ceolin
User Modelingit is not about putting everything in a user profile
it is about making the right choices
Social Web 2016 Lora Aroyo and Davide Ceolin
User Adaptation
Knowing the user to adapt a system or interfaceto improve the system functionality and user
experience
Social Web 2016 Lora Aroyo and Davide Ceolin
A Jameson Adaptive interfaces and agents The HCI handbook fundamentals evolving technologies and emerging applications pp 305ndash330 2003
User-Adaptive Systems
Social Web 2016 Lora Aroyo and Davide Ceolin
based on slides from Fabien Abel
Lastfm adapts to your music taste
bull Overfitting ldquobubble effectsrdquo loss of serendipity problem bull systems may adapt too strongly to the
interestsbehaviorbull eg an adaptive radio station may always play the
same or very similar songsbull We search for the right balance between novelty and
relevance for the userbull ldquoLost in Hyperspacerdquo problem
bull when adapting the navigation ndash ie the links on which users can click to findaccess information
bull eg re-orderinghiding of menu items may lead to confusion
Issues in User-Adaptive Systems
Social Web 2016 Lora Aroyo and Davide Ceolin
What is good user modelling amp
personalisation
httpwwwflickrcomphotosbellarosebyliz4729613108
From the consumer perspective of an adaptive system
From the provider perspective of an adaptive system
Success Perspectives
Social Web 2016 Lora Aroyo and Davide Ceolin
bull User studies askobserve (selected) people whether you did a good job
bull Log analysis Analyze (click) data and infer whether you did a good job
bull Evaluation of user modelingbull measure quality of profiles directly eg measure
overlap with existing (true) profiles or let people judge the quality of the generated user profiles
bull measure quality of application that exploits the user profile eg apply user modeling strategies in a recommender system
Evaluation Strategies
Social Web 2016 Lora Aroyo and Davide Ceolin
Evaluating User Modeling in RecSys
Social Web 2016 Lora Aroyo and Davide Ceolin
Possible Metrics
Social Web 2016 Lora Aroyo and Davide Ceolin
bull The usual IR metricsbull Precision fraction of retrieved items that are relevantbull Recall fraction of relevant items that have been
retrievedbull F-Measure (harmonic) mean of precision and recall
bull Metrics for evaluating recommendation (rankings)bull Mean Reciprocal Rank (MRR) of first relevant itembull Successk probability that relevant item occurs
within the top kbull If a true ranking is given rank correlations bull Precisionk Recallk amp F-Measurek
bull Metrics for evaluating prediction of user preferencesbull MAE = Mean Absolute Errorbull TrueFalse PositivesNegatives
Predict relevantusefulinteresting items for a given user (in a given context) itrsquos often a ranking task
Recommendation Systems
Social Web 2016 Lora Aroyo and Davide Ceolin
March 28 2013
Social Web 2016 Lora Aroyo and Davide Ceolin
Social Web 2016 Lora Aroyo and Davide Ceolin
Social Web 2016 Lora Aroyo and Davide Ceolin
httpwwwwiredcommagazine201111mf_artsyall1
Social Web 2016 Lora Aroyo and Davide Ceolin
Collaborative Filteringbull Memory-based User-Item matrix ratingspreferences of users =gt
compute similarity between users amp recommend items of similar users
bull Model-based Item-Item matrix similarity (eg based on user ratings) between items =gt recommend items that are similar to the ones the user likes
bull Model-based Clustering cluster users according to their preferences =gt recommend items of users that belong to the same cluster
bull Model-based Bayesian networks P(u likes item B | u likes item A) = how likely is it that a user who likes item A will like item B learn probabilities from user ratingspreferences
bull Others rule-based other data mining techniques
Social Web 2016 Lora Aroyo and Davide Ceolin
bull complete input data is required
bull pre-computation not possible
bull does not scale well bull high quality of
recommendations
bull abstraction (model) of input data
bull pre-computation (partially) possible (model has to be re-built from time to time)
bull scales betterbull abstraction may reduce
recommendation quality
Memory vs Model-based
Social Web 2016 Lora Aroyo and Davide Ceolin
bull collaborative filtering lsquoneighborhoodsrsquo of people with similar interest amp recommending items based on likings in neighborhood
bull limitations next to lsquocold startrsquo and lsquosparsityrsquo the lack of control (over onersquos neighborhood) is also a problem ie cannot add lsquotrustedrsquo people nor exclude lsquostrangersquo ones
bull therefore interest in lsquosocial recommendersrsquo where presence of social connections defines the similarity in interests (eg social tagging CiteULike)bull does a social connection indicate user interest similaritybull how much users interest similarity depends on the strength
of their connectionbull is it feasible to use a social network as a personalized
recommendation[Lin amp Brusilovsky Social Networks and Interest Similarity The Case of CiteULike HTrsquo10]
Social Networks amp Interest Similarity
bull unilaterally connected pairs have more common itemsmetadatatags than non-connected pairsbull highest similarity for direct connections - decreasing with the increase of distance between users in SN bull reciprocal relationship users - significantly larger similarity than users in a unidirectional relationship bull traditional item-level similarity may be less reliable to find similar users in social bookmarking systemsbull peers connected by self-defined social connections could be a useful source for cross-recommendation
Conclusions
Social Web 2016 Lora Aroyo and Davide Ceolin
bull Input characteristics of items amp interests of a user into characteristics of items =gt Recommend items that feature characteristics which meet the userrsquos interests
bull Techniquesbull Data mining methods Cluster items based on their
characteristics =gt Infer usersrsquo interests into clustersbull IR methods Represent items amp users as term vectors
=gt Compute similarity between user profile vector and items
bull Utility-based methods Utility function that gets an item as input the parameters of the utility function are customized via preferences of a user
Content-based Recommendations
Social Web 2016 Lora Aroyo and Davide Ceolin
based on slides from Fabien Abel
Content Features
based on slides from Fabien Abel
User Model
based on slides from Fabien Abel
Recommendations
RecSys Issuesbull Cold-start problem (new user problem) nolittle data available to infer preferences of new
usersbull Changing User Preferences user interests may change over timebull Sparsity problem (new item problem) item descriptions are sparse eg not many user
rated or tagged an itembull Lack of Diversity (overfitting) when adapting too strongly to the preferences of users they
might see samesimilar recommendationsbull Use the right context users do things which might not be relevant for their user model
eg try out things do stuff for other peoplebull Research challenge right balance between serendipity amp personalizationbull Research challenge right way to use the influence of recommendations on userrsquos behavior
Social Web 2016 Lora Aroyo and Davide Ceolin
Announcementsbull Next deadlinesbull Tuesday March 1st 2359
Assignment 2bull Friday March 4th 1000 Post
questionbull Friday March 4th 1700 Vote
questions
image source httpwwwflickrcomphotosbionicteaching1375254387
Hands-on Teaser
bull Your Facebook Friendsrsquo popularity in a spread sheet
bull Locations of your Facebook Friendsbull Tag Cloud of your wall posts
- Social Web 2016
- Slide 2
- Slide 3
- Slide 4
- Slide 5
- Slide 6
- Slide 7
- Slide 8
- Slide 9
- Slide 10
- Slide 11
- Slide 12
- Slide 13
- Slide 14
- Slide 15
- Slide 16
- Slide 17
- Slide 18
- Slide 19
- Slide 20
- Slide 21
- Slide 22
- Slide 23
- Slide 24
- Slide 25
- Slide 26
- Slide 27
- Slide 28
- Slide 29
- Slide 30
- Slide 31
- Slide 32
- Slide 33
- Slide 34
- Slide 35
- Slide 36
- Slide 37
- Slide 38
- Slide 39
- Slide 40
- Slide 41
- Slide 42
- Slide 43
- Slide 44
- Slide 45
- Slide 46
- Slide 47
- Slide 48
- Announcements
- Slide 50
-
Fig 1 Functional model of tasks and sub-tasks specifically suited for SASs
Functional model of tasks and sub-tasks specifically suited for SASs (Ilaria Torre 2009)
Social Web 2016 Lora Aroyo and Davide Ceolin
Kevin Kelly
How to infer amp represent user information that supports a given application or context
User Modeling
Social Web 2016 Lora Aroyo and Davide Ceolin
bull Application has to obtain understand amp exploit information about the user
bull Information (need amp context) about user
bull Inferring information about user amp representing it so that it can be consumed by the application
bull Data relevant for inferring information about user
User Modeling Challenge
Social Web 2016 Lora Aroyo and Davide Ceolin
bull People leave traces on the Web and on their computersbull Usage data eg query logs click-through-data bull Social data eg tags (micro-)blog posts comments
bookmarks friend connections bull Documents eg pictures videosbull Personal data eg affiliations locations bull Products applications services - bought used installed
bull Not only a userrsquos behavior but also interactions of other users bull ldquopeople can make statements about merdquobull ldquopeople who are similar to me can reveal information
about merdquobull ldquosocial learningrdquo collaborative recommender systems
User amp Usage Datais Everywhere
Social Web 2016 Lora Aroyo and Davide Ceolin
bull User Profile = data structure = a characterization of a user at a particular moment represents what from a given (system) perspective there is to know about a user The data in the profile can be explicitly given by user or derived by system
bull User Model = definitions amp rules for the interpretation of observations about the user amp about the translation of that interpretation into the characteristics in a user profile user model is the recipe for obtaining amp interpreting user profiles
bull User Modeling = the process of representing the user
UM Basic Concepts
Social Web 2016 Lora Aroyo and Davide Ceolin
bull Overlay User Modeling describe user characteristics eg ldquoknowledge of a userrdquo ldquointerests of a userrdquo with respect to ldquoidealrdquo characteristics
bull Customizing user explicitly provides amp adjusts elements of the user profile
bull User model elicitation ask amp observe the user learn amp improve user profile successively ldquointeractive user modelingrdquo
bull Stereotyping stereotypical characteristics to describe a user
bull User Relevance Modeling learninfer probabilities that a given item or concept is relevant for a user
Related scientific conference httpumap2011org Related journal httpumuaiorg
User Modeling Approaches
Social Web 2016 Lora Aroyo and Davide Ceolin
httpfarm7staticflickrcom62406346803873_e756dd9bae_bjpg
Which approach suits best the conditions of
applications
bull among the oldest user modelsbull used for modeling student
knowledgebull the user is typically
characterized in terms of domain concepts amp hypotheses of the userrsquos knowledge about these concepts in relation to an (ideal) expertrsquos knowledge
bull concept-value pairs
Overlay User Models
Social Web 2016 Lora Aroyo and Davide Ceolin
bull Ask the user explicitly learnbull NLP intelligent dialoguesbull Bayesian networks Hidden Markov models
bull Observe the user learn bull Logs machine learningbull Clustering classification data mining
bull Interactive user modeling mixture of direct inputs of a user observations and inferences
User Model Elicitation
Social Web 2016 Lora Aroyo and Davide Ceolin
httphunchcomSocial Web 2016 Lora Aroyo and Davide Ceolin
Social Web 2016 Lora Aroyo and Davide Ceolin
httpfarm1staticflickrcom155413650229_31ef379b0b_bjpg
bull set of characteristics (eg attribute-value pairs) that describe a group of users
bull user is not assigned to a single stereotype - user profile can feature characteristics of several different stereotypes
User Stereotypes
based on slides from Fabien Abel
based on slides from Fabien Abel
Can we infer a Twitter-based User
Profile
User Modeling Building Blocks
based on slides from Fabien Abel
User Modeling Building Blocks
based on slides from Fabien Abel
User Modeling Building Blocks
based on slides from Fabien Abel
User Modeling Building Blocks
based on slides from Fabien Abel
Observationsbull Profile characteristics
bull Semantic enrichment solves sparsity problemsbull Profiles change over time recent profiles reflect
better current user demandsbull Temporal patterns weekend profiles differ
significantly from weekday profilesbull Impact on recommendations
bull The more fine-grained the concepts the better the recommendation performance entity-based gt topic-based gt hashtag-based
bull Semantic enrichment improves recommendation quality
bull Time-sensitivity (adapting to trends) improves performance
Social Web 2016 Lora Aroyo and Davide Ceolin
User Modelingit is not about putting everything in a user profile
it is about making the right choices
Social Web 2016 Lora Aroyo and Davide Ceolin
User Adaptation
Knowing the user to adapt a system or interfaceto improve the system functionality and user
experience
Social Web 2016 Lora Aroyo and Davide Ceolin
A Jameson Adaptive interfaces and agents The HCI handbook fundamentals evolving technologies and emerging applications pp 305ndash330 2003
User-Adaptive Systems
Social Web 2016 Lora Aroyo and Davide Ceolin
based on slides from Fabien Abel
Lastfm adapts to your music taste
bull Overfitting ldquobubble effectsrdquo loss of serendipity problem bull systems may adapt too strongly to the
interestsbehaviorbull eg an adaptive radio station may always play the
same or very similar songsbull We search for the right balance between novelty and
relevance for the userbull ldquoLost in Hyperspacerdquo problem
bull when adapting the navigation ndash ie the links on which users can click to findaccess information
bull eg re-orderinghiding of menu items may lead to confusion
Issues in User-Adaptive Systems
Social Web 2016 Lora Aroyo and Davide Ceolin
What is good user modelling amp
personalisation
httpwwwflickrcomphotosbellarosebyliz4729613108
From the consumer perspective of an adaptive system
From the provider perspective of an adaptive system
Success Perspectives
Social Web 2016 Lora Aroyo and Davide Ceolin
bull User studies askobserve (selected) people whether you did a good job
bull Log analysis Analyze (click) data and infer whether you did a good job
bull Evaluation of user modelingbull measure quality of profiles directly eg measure
overlap with existing (true) profiles or let people judge the quality of the generated user profiles
bull measure quality of application that exploits the user profile eg apply user modeling strategies in a recommender system
Evaluation Strategies
Social Web 2016 Lora Aroyo and Davide Ceolin
Evaluating User Modeling in RecSys
Social Web 2016 Lora Aroyo and Davide Ceolin
Possible Metrics
Social Web 2016 Lora Aroyo and Davide Ceolin
bull The usual IR metricsbull Precision fraction of retrieved items that are relevantbull Recall fraction of relevant items that have been
retrievedbull F-Measure (harmonic) mean of precision and recall
bull Metrics for evaluating recommendation (rankings)bull Mean Reciprocal Rank (MRR) of first relevant itembull Successk probability that relevant item occurs
within the top kbull If a true ranking is given rank correlations bull Precisionk Recallk amp F-Measurek
bull Metrics for evaluating prediction of user preferencesbull MAE = Mean Absolute Errorbull TrueFalse PositivesNegatives
Predict relevantusefulinteresting items for a given user (in a given context) itrsquos often a ranking task
Recommendation Systems
Social Web 2016 Lora Aroyo and Davide Ceolin
March 28 2013
Social Web 2016 Lora Aroyo and Davide Ceolin
Social Web 2016 Lora Aroyo and Davide Ceolin
Social Web 2016 Lora Aroyo and Davide Ceolin
httpwwwwiredcommagazine201111mf_artsyall1
Social Web 2016 Lora Aroyo and Davide Ceolin
Collaborative Filteringbull Memory-based User-Item matrix ratingspreferences of users =gt
compute similarity between users amp recommend items of similar users
bull Model-based Item-Item matrix similarity (eg based on user ratings) between items =gt recommend items that are similar to the ones the user likes
bull Model-based Clustering cluster users according to their preferences =gt recommend items of users that belong to the same cluster
bull Model-based Bayesian networks P(u likes item B | u likes item A) = how likely is it that a user who likes item A will like item B learn probabilities from user ratingspreferences
bull Others rule-based other data mining techniques
Social Web 2016 Lora Aroyo and Davide Ceolin
bull complete input data is required
bull pre-computation not possible
bull does not scale well bull high quality of
recommendations
bull abstraction (model) of input data
bull pre-computation (partially) possible (model has to be re-built from time to time)
bull scales betterbull abstraction may reduce
recommendation quality
Memory vs Model-based
Social Web 2016 Lora Aroyo and Davide Ceolin
bull collaborative filtering lsquoneighborhoodsrsquo of people with similar interest amp recommending items based on likings in neighborhood
bull limitations next to lsquocold startrsquo and lsquosparsityrsquo the lack of control (over onersquos neighborhood) is also a problem ie cannot add lsquotrustedrsquo people nor exclude lsquostrangersquo ones
bull therefore interest in lsquosocial recommendersrsquo where presence of social connections defines the similarity in interests (eg social tagging CiteULike)bull does a social connection indicate user interest similaritybull how much users interest similarity depends on the strength
of their connectionbull is it feasible to use a social network as a personalized
recommendation[Lin amp Brusilovsky Social Networks and Interest Similarity The Case of CiteULike HTrsquo10]
Social Networks amp Interest Similarity
bull unilaterally connected pairs have more common itemsmetadatatags than non-connected pairsbull highest similarity for direct connections - decreasing with the increase of distance between users in SN bull reciprocal relationship users - significantly larger similarity than users in a unidirectional relationship bull traditional item-level similarity may be less reliable to find similar users in social bookmarking systemsbull peers connected by self-defined social connections could be a useful source for cross-recommendation
Conclusions
Social Web 2016 Lora Aroyo and Davide Ceolin
bull Input characteristics of items amp interests of a user into characteristics of items =gt Recommend items that feature characteristics which meet the userrsquos interests
bull Techniquesbull Data mining methods Cluster items based on their
characteristics =gt Infer usersrsquo interests into clustersbull IR methods Represent items amp users as term vectors
=gt Compute similarity between user profile vector and items
bull Utility-based methods Utility function that gets an item as input the parameters of the utility function are customized via preferences of a user
Content-based Recommendations
Social Web 2016 Lora Aroyo and Davide Ceolin
based on slides from Fabien Abel
Content Features
based on slides from Fabien Abel
User Model
based on slides from Fabien Abel
Recommendations
RecSys Issuesbull Cold-start problem (new user problem) nolittle data available to infer preferences of new
usersbull Changing User Preferences user interests may change over timebull Sparsity problem (new item problem) item descriptions are sparse eg not many user
rated or tagged an itembull Lack of Diversity (overfitting) when adapting too strongly to the preferences of users they
might see samesimilar recommendationsbull Use the right context users do things which might not be relevant for their user model
eg try out things do stuff for other peoplebull Research challenge right balance between serendipity amp personalizationbull Research challenge right way to use the influence of recommendations on userrsquos behavior
Social Web 2016 Lora Aroyo and Davide Ceolin
Announcementsbull Next deadlinesbull Tuesday March 1st 2359
Assignment 2bull Friday March 4th 1000 Post
questionbull Friday March 4th 1700 Vote
questions
image source httpwwwflickrcomphotosbionicteaching1375254387
Hands-on Teaser
bull Your Facebook Friendsrsquo popularity in a spread sheet
bull Locations of your Facebook Friendsbull Tag Cloud of your wall posts
- Social Web 2016
- Slide 2
- Slide 3
- Slide 4
- Slide 5
- Slide 6
- Slide 7
- Slide 8
- Slide 9
- Slide 10
- Slide 11
- Slide 12
- Slide 13
- Slide 14
- Slide 15
- Slide 16
- Slide 17
- Slide 18
- Slide 19
- Slide 20
- Slide 21
- Slide 22
- Slide 23
- Slide 24
- Slide 25
- Slide 26
- Slide 27
- Slide 28
- Slide 29
- Slide 30
- Slide 31
- Slide 32
- Slide 33
- Slide 34
- Slide 35
- Slide 36
- Slide 37
- Slide 38
- Slide 39
- Slide 40
- Slide 41
- Slide 42
- Slide 43
- Slide 44
- Slide 45
- Slide 46
- Slide 47
- Slide 48
- Announcements
- Slide 50
-
Kevin Kelly
How to infer amp represent user information that supports a given application or context
User Modeling
Social Web 2016 Lora Aroyo and Davide Ceolin
bull Application has to obtain understand amp exploit information about the user
bull Information (need amp context) about user
bull Inferring information about user amp representing it so that it can be consumed by the application
bull Data relevant for inferring information about user
User Modeling Challenge
Social Web 2016 Lora Aroyo and Davide Ceolin
bull People leave traces on the Web and on their computersbull Usage data eg query logs click-through-data bull Social data eg tags (micro-)blog posts comments
bookmarks friend connections bull Documents eg pictures videosbull Personal data eg affiliations locations bull Products applications services - bought used installed
bull Not only a userrsquos behavior but also interactions of other users bull ldquopeople can make statements about merdquobull ldquopeople who are similar to me can reveal information
about merdquobull ldquosocial learningrdquo collaborative recommender systems
User amp Usage Datais Everywhere
Social Web 2016 Lora Aroyo and Davide Ceolin
bull User Profile = data structure = a characterization of a user at a particular moment represents what from a given (system) perspective there is to know about a user The data in the profile can be explicitly given by user or derived by system
bull User Model = definitions amp rules for the interpretation of observations about the user amp about the translation of that interpretation into the characteristics in a user profile user model is the recipe for obtaining amp interpreting user profiles
bull User Modeling = the process of representing the user
UM Basic Concepts
Social Web 2016 Lora Aroyo and Davide Ceolin
bull Overlay User Modeling describe user characteristics eg ldquoknowledge of a userrdquo ldquointerests of a userrdquo with respect to ldquoidealrdquo characteristics
bull Customizing user explicitly provides amp adjusts elements of the user profile
bull User model elicitation ask amp observe the user learn amp improve user profile successively ldquointeractive user modelingrdquo
bull Stereotyping stereotypical characteristics to describe a user
bull User Relevance Modeling learninfer probabilities that a given item or concept is relevant for a user
Related scientific conference httpumap2011org Related journal httpumuaiorg
User Modeling Approaches
Social Web 2016 Lora Aroyo and Davide Ceolin
httpfarm7staticflickrcom62406346803873_e756dd9bae_bjpg
Which approach suits best the conditions of
applications
bull among the oldest user modelsbull used for modeling student
knowledgebull the user is typically
characterized in terms of domain concepts amp hypotheses of the userrsquos knowledge about these concepts in relation to an (ideal) expertrsquos knowledge
bull concept-value pairs
Overlay User Models
Social Web 2016 Lora Aroyo and Davide Ceolin
bull Ask the user explicitly learnbull NLP intelligent dialoguesbull Bayesian networks Hidden Markov models
bull Observe the user learn bull Logs machine learningbull Clustering classification data mining
bull Interactive user modeling mixture of direct inputs of a user observations and inferences
User Model Elicitation
Social Web 2016 Lora Aroyo and Davide Ceolin
httphunchcomSocial Web 2016 Lora Aroyo and Davide Ceolin
Social Web 2016 Lora Aroyo and Davide Ceolin
httpfarm1staticflickrcom155413650229_31ef379b0b_bjpg
bull set of characteristics (eg attribute-value pairs) that describe a group of users
bull user is not assigned to a single stereotype - user profile can feature characteristics of several different stereotypes
User Stereotypes
based on slides from Fabien Abel
based on slides from Fabien Abel
Can we infer a Twitter-based User
Profile
User Modeling Building Blocks
based on slides from Fabien Abel
User Modeling Building Blocks
based on slides from Fabien Abel
User Modeling Building Blocks
based on slides from Fabien Abel
User Modeling Building Blocks
based on slides from Fabien Abel
Observationsbull Profile characteristics
bull Semantic enrichment solves sparsity problemsbull Profiles change over time recent profiles reflect
better current user demandsbull Temporal patterns weekend profiles differ
significantly from weekday profilesbull Impact on recommendations
bull The more fine-grained the concepts the better the recommendation performance entity-based gt topic-based gt hashtag-based
bull Semantic enrichment improves recommendation quality
bull Time-sensitivity (adapting to trends) improves performance
Social Web 2016 Lora Aroyo and Davide Ceolin
User Modelingit is not about putting everything in a user profile
it is about making the right choices
Social Web 2016 Lora Aroyo and Davide Ceolin
User Adaptation
Knowing the user to adapt a system or interfaceto improve the system functionality and user
experience
Social Web 2016 Lora Aroyo and Davide Ceolin
A Jameson Adaptive interfaces and agents The HCI handbook fundamentals evolving technologies and emerging applications pp 305ndash330 2003
User-Adaptive Systems
Social Web 2016 Lora Aroyo and Davide Ceolin
based on slides from Fabien Abel
Lastfm adapts to your music taste
bull Overfitting ldquobubble effectsrdquo loss of serendipity problem bull systems may adapt too strongly to the
interestsbehaviorbull eg an adaptive radio station may always play the
same or very similar songsbull We search for the right balance between novelty and
relevance for the userbull ldquoLost in Hyperspacerdquo problem
bull when adapting the navigation ndash ie the links on which users can click to findaccess information
bull eg re-orderinghiding of menu items may lead to confusion
Issues in User-Adaptive Systems
Social Web 2016 Lora Aroyo and Davide Ceolin
What is good user modelling amp
personalisation
httpwwwflickrcomphotosbellarosebyliz4729613108
From the consumer perspective of an adaptive system
From the provider perspective of an adaptive system
Success Perspectives
Social Web 2016 Lora Aroyo and Davide Ceolin
bull User studies askobserve (selected) people whether you did a good job
bull Log analysis Analyze (click) data and infer whether you did a good job
bull Evaluation of user modelingbull measure quality of profiles directly eg measure
overlap with existing (true) profiles or let people judge the quality of the generated user profiles
bull measure quality of application that exploits the user profile eg apply user modeling strategies in a recommender system
Evaluation Strategies
Social Web 2016 Lora Aroyo and Davide Ceolin
Evaluating User Modeling in RecSys
Social Web 2016 Lora Aroyo and Davide Ceolin
Possible Metrics
Social Web 2016 Lora Aroyo and Davide Ceolin
bull The usual IR metricsbull Precision fraction of retrieved items that are relevantbull Recall fraction of relevant items that have been
retrievedbull F-Measure (harmonic) mean of precision and recall
bull Metrics for evaluating recommendation (rankings)bull Mean Reciprocal Rank (MRR) of first relevant itembull Successk probability that relevant item occurs
within the top kbull If a true ranking is given rank correlations bull Precisionk Recallk amp F-Measurek
bull Metrics for evaluating prediction of user preferencesbull MAE = Mean Absolute Errorbull TrueFalse PositivesNegatives
Predict relevantusefulinteresting items for a given user (in a given context) itrsquos often a ranking task
Recommendation Systems
Social Web 2016 Lora Aroyo and Davide Ceolin
March 28 2013
Social Web 2016 Lora Aroyo and Davide Ceolin
Social Web 2016 Lora Aroyo and Davide Ceolin
Social Web 2016 Lora Aroyo and Davide Ceolin
httpwwwwiredcommagazine201111mf_artsyall1
Social Web 2016 Lora Aroyo and Davide Ceolin
Collaborative Filteringbull Memory-based User-Item matrix ratingspreferences of users =gt
compute similarity between users amp recommend items of similar users
bull Model-based Item-Item matrix similarity (eg based on user ratings) between items =gt recommend items that are similar to the ones the user likes
bull Model-based Clustering cluster users according to their preferences =gt recommend items of users that belong to the same cluster
bull Model-based Bayesian networks P(u likes item B | u likes item A) = how likely is it that a user who likes item A will like item B learn probabilities from user ratingspreferences
bull Others rule-based other data mining techniques
Social Web 2016 Lora Aroyo and Davide Ceolin
bull complete input data is required
bull pre-computation not possible
bull does not scale well bull high quality of
recommendations
bull abstraction (model) of input data
bull pre-computation (partially) possible (model has to be re-built from time to time)
bull scales betterbull abstraction may reduce
recommendation quality
Memory vs Model-based
Social Web 2016 Lora Aroyo and Davide Ceolin
bull collaborative filtering lsquoneighborhoodsrsquo of people with similar interest amp recommending items based on likings in neighborhood
bull limitations next to lsquocold startrsquo and lsquosparsityrsquo the lack of control (over onersquos neighborhood) is also a problem ie cannot add lsquotrustedrsquo people nor exclude lsquostrangersquo ones
bull therefore interest in lsquosocial recommendersrsquo where presence of social connections defines the similarity in interests (eg social tagging CiteULike)bull does a social connection indicate user interest similaritybull how much users interest similarity depends on the strength
of their connectionbull is it feasible to use a social network as a personalized
recommendation[Lin amp Brusilovsky Social Networks and Interest Similarity The Case of CiteULike HTrsquo10]
Social Networks amp Interest Similarity
bull unilaterally connected pairs have more common itemsmetadatatags than non-connected pairsbull highest similarity for direct connections - decreasing with the increase of distance between users in SN bull reciprocal relationship users - significantly larger similarity than users in a unidirectional relationship bull traditional item-level similarity may be less reliable to find similar users in social bookmarking systemsbull peers connected by self-defined social connections could be a useful source for cross-recommendation
Conclusions
Social Web 2016 Lora Aroyo and Davide Ceolin
bull Input characteristics of items amp interests of a user into characteristics of items =gt Recommend items that feature characteristics which meet the userrsquos interests
bull Techniquesbull Data mining methods Cluster items based on their
characteristics =gt Infer usersrsquo interests into clustersbull IR methods Represent items amp users as term vectors
=gt Compute similarity between user profile vector and items
bull Utility-based methods Utility function that gets an item as input the parameters of the utility function are customized via preferences of a user
Content-based Recommendations
Social Web 2016 Lora Aroyo and Davide Ceolin
based on slides from Fabien Abel
Content Features
based on slides from Fabien Abel
User Model
based on slides from Fabien Abel
Recommendations
RecSys Issuesbull Cold-start problem (new user problem) nolittle data available to infer preferences of new
usersbull Changing User Preferences user interests may change over timebull Sparsity problem (new item problem) item descriptions are sparse eg not many user
rated or tagged an itembull Lack of Diversity (overfitting) when adapting too strongly to the preferences of users they
might see samesimilar recommendationsbull Use the right context users do things which might not be relevant for their user model
eg try out things do stuff for other peoplebull Research challenge right balance between serendipity amp personalizationbull Research challenge right way to use the influence of recommendations on userrsquos behavior
Social Web 2016 Lora Aroyo and Davide Ceolin
Announcementsbull Next deadlinesbull Tuesday March 1st 2359
Assignment 2bull Friday March 4th 1000 Post
questionbull Friday March 4th 1700 Vote
questions
image source httpwwwflickrcomphotosbionicteaching1375254387
Hands-on Teaser
bull Your Facebook Friendsrsquo popularity in a spread sheet
bull Locations of your Facebook Friendsbull Tag Cloud of your wall posts
- Social Web 2016
- Slide 2
- Slide 3
- Slide 4
- Slide 5
- Slide 6
- Slide 7
- Slide 8
- Slide 9
- Slide 10
- Slide 11
- Slide 12
- Slide 13
- Slide 14
- Slide 15
- Slide 16
- Slide 17
- Slide 18
- Slide 19
- Slide 20
- Slide 21
- Slide 22
- Slide 23
- Slide 24
- Slide 25
- Slide 26
- Slide 27
- Slide 28
- Slide 29
- Slide 30
- Slide 31
- Slide 32
- Slide 33
- Slide 34
- Slide 35
- Slide 36
- Slide 37
- Slide 38
- Slide 39
- Slide 40
- Slide 41
- Slide 42
- Slide 43
- Slide 44
- Slide 45
- Slide 46
- Slide 47
- Slide 48
- Announcements
- Slide 50
-
bull Application has to obtain understand amp exploit information about the user
bull Information (need amp context) about user
bull Inferring information about user amp representing it so that it can be consumed by the application
bull Data relevant for inferring information about user
User Modeling Challenge
Social Web 2016 Lora Aroyo and Davide Ceolin
bull People leave traces on the Web and on their computersbull Usage data eg query logs click-through-data bull Social data eg tags (micro-)blog posts comments
bookmarks friend connections bull Documents eg pictures videosbull Personal data eg affiliations locations bull Products applications services - bought used installed
bull Not only a userrsquos behavior but also interactions of other users bull ldquopeople can make statements about merdquobull ldquopeople who are similar to me can reveal information
about merdquobull ldquosocial learningrdquo collaborative recommender systems
User amp Usage Datais Everywhere
Social Web 2016 Lora Aroyo and Davide Ceolin
bull User Profile = data structure = a characterization of a user at a particular moment represents what from a given (system) perspective there is to know about a user The data in the profile can be explicitly given by user or derived by system
bull User Model = definitions amp rules for the interpretation of observations about the user amp about the translation of that interpretation into the characteristics in a user profile user model is the recipe for obtaining amp interpreting user profiles
bull User Modeling = the process of representing the user
UM Basic Concepts
Social Web 2016 Lora Aroyo and Davide Ceolin
bull Overlay User Modeling describe user characteristics eg ldquoknowledge of a userrdquo ldquointerests of a userrdquo with respect to ldquoidealrdquo characteristics
bull Customizing user explicitly provides amp adjusts elements of the user profile
bull User model elicitation ask amp observe the user learn amp improve user profile successively ldquointeractive user modelingrdquo
bull Stereotyping stereotypical characteristics to describe a user
bull User Relevance Modeling learninfer probabilities that a given item or concept is relevant for a user
Related scientific conference httpumap2011org Related journal httpumuaiorg
User Modeling Approaches
Social Web 2016 Lora Aroyo and Davide Ceolin
httpfarm7staticflickrcom62406346803873_e756dd9bae_bjpg
Which approach suits best the conditions of
applications
bull among the oldest user modelsbull used for modeling student
knowledgebull the user is typically
characterized in terms of domain concepts amp hypotheses of the userrsquos knowledge about these concepts in relation to an (ideal) expertrsquos knowledge
bull concept-value pairs
Overlay User Models
Social Web 2016 Lora Aroyo and Davide Ceolin
bull Ask the user explicitly learnbull NLP intelligent dialoguesbull Bayesian networks Hidden Markov models
bull Observe the user learn bull Logs machine learningbull Clustering classification data mining
bull Interactive user modeling mixture of direct inputs of a user observations and inferences
User Model Elicitation
Social Web 2016 Lora Aroyo and Davide Ceolin
httphunchcomSocial Web 2016 Lora Aroyo and Davide Ceolin
Social Web 2016 Lora Aroyo and Davide Ceolin
httpfarm1staticflickrcom155413650229_31ef379b0b_bjpg
bull set of characteristics (eg attribute-value pairs) that describe a group of users
bull user is not assigned to a single stereotype - user profile can feature characteristics of several different stereotypes
User Stereotypes
based on slides from Fabien Abel
based on slides from Fabien Abel
Can we infer a Twitter-based User
Profile
User Modeling Building Blocks
based on slides from Fabien Abel
User Modeling Building Blocks
based on slides from Fabien Abel
User Modeling Building Blocks
based on slides from Fabien Abel
User Modeling Building Blocks
based on slides from Fabien Abel
Observationsbull Profile characteristics
bull Semantic enrichment solves sparsity problemsbull Profiles change over time recent profiles reflect
better current user demandsbull Temporal patterns weekend profiles differ
significantly from weekday profilesbull Impact on recommendations
bull The more fine-grained the concepts the better the recommendation performance entity-based gt topic-based gt hashtag-based
bull Semantic enrichment improves recommendation quality
bull Time-sensitivity (adapting to trends) improves performance
Social Web 2016 Lora Aroyo and Davide Ceolin
User Modelingit is not about putting everything in a user profile
it is about making the right choices
Social Web 2016 Lora Aroyo and Davide Ceolin
User Adaptation
Knowing the user to adapt a system or interfaceto improve the system functionality and user
experience
Social Web 2016 Lora Aroyo and Davide Ceolin
A Jameson Adaptive interfaces and agents The HCI handbook fundamentals evolving technologies and emerging applications pp 305ndash330 2003
User-Adaptive Systems
Social Web 2016 Lora Aroyo and Davide Ceolin
based on slides from Fabien Abel
Lastfm adapts to your music taste
bull Overfitting ldquobubble effectsrdquo loss of serendipity problem bull systems may adapt too strongly to the
interestsbehaviorbull eg an adaptive radio station may always play the
same or very similar songsbull We search for the right balance between novelty and
relevance for the userbull ldquoLost in Hyperspacerdquo problem
bull when adapting the navigation ndash ie the links on which users can click to findaccess information
bull eg re-orderinghiding of menu items may lead to confusion
Issues in User-Adaptive Systems
Social Web 2016 Lora Aroyo and Davide Ceolin
What is good user modelling amp
personalisation
httpwwwflickrcomphotosbellarosebyliz4729613108
From the consumer perspective of an adaptive system
From the provider perspective of an adaptive system
Success Perspectives
Social Web 2016 Lora Aroyo and Davide Ceolin
bull User studies askobserve (selected) people whether you did a good job
bull Log analysis Analyze (click) data and infer whether you did a good job
bull Evaluation of user modelingbull measure quality of profiles directly eg measure
overlap with existing (true) profiles or let people judge the quality of the generated user profiles
bull measure quality of application that exploits the user profile eg apply user modeling strategies in a recommender system
Evaluation Strategies
Social Web 2016 Lora Aroyo and Davide Ceolin
Evaluating User Modeling in RecSys
Social Web 2016 Lora Aroyo and Davide Ceolin
Possible Metrics
Social Web 2016 Lora Aroyo and Davide Ceolin
bull The usual IR metricsbull Precision fraction of retrieved items that are relevantbull Recall fraction of relevant items that have been
retrievedbull F-Measure (harmonic) mean of precision and recall
bull Metrics for evaluating recommendation (rankings)bull Mean Reciprocal Rank (MRR) of first relevant itembull Successk probability that relevant item occurs
within the top kbull If a true ranking is given rank correlations bull Precisionk Recallk amp F-Measurek
bull Metrics for evaluating prediction of user preferencesbull MAE = Mean Absolute Errorbull TrueFalse PositivesNegatives
Predict relevantusefulinteresting items for a given user (in a given context) itrsquos often a ranking task
Recommendation Systems
Social Web 2016 Lora Aroyo and Davide Ceolin
March 28 2013
Social Web 2016 Lora Aroyo and Davide Ceolin
Social Web 2016 Lora Aroyo and Davide Ceolin
Social Web 2016 Lora Aroyo and Davide Ceolin
httpwwwwiredcommagazine201111mf_artsyall1
Social Web 2016 Lora Aroyo and Davide Ceolin
Collaborative Filteringbull Memory-based User-Item matrix ratingspreferences of users =gt
compute similarity between users amp recommend items of similar users
bull Model-based Item-Item matrix similarity (eg based on user ratings) between items =gt recommend items that are similar to the ones the user likes
bull Model-based Clustering cluster users according to their preferences =gt recommend items of users that belong to the same cluster
bull Model-based Bayesian networks P(u likes item B | u likes item A) = how likely is it that a user who likes item A will like item B learn probabilities from user ratingspreferences
bull Others rule-based other data mining techniques
Social Web 2016 Lora Aroyo and Davide Ceolin
bull complete input data is required
bull pre-computation not possible
bull does not scale well bull high quality of
recommendations
bull abstraction (model) of input data
bull pre-computation (partially) possible (model has to be re-built from time to time)
bull scales betterbull abstraction may reduce
recommendation quality
Memory vs Model-based
Social Web 2016 Lora Aroyo and Davide Ceolin
bull collaborative filtering lsquoneighborhoodsrsquo of people with similar interest amp recommending items based on likings in neighborhood
bull limitations next to lsquocold startrsquo and lsquosparsityrsquo the lack of control (over onersquos neighborhood) is also a problem ie cannot add lsquotrustedrsquo people nor exclude lsquostrangersquo ones
bull therefore interest in lsquosocial recommendersrsquo where presence of social connections defines the similarity in interests (eg social tagging CiteULike)bull does a social connection indicate user interest similaritybull how much users interest similarity depends on the strength
of their connectionbull is it feasible to use a social network as a personalized
recommendation[Lin amp Brusilovsky Social Networks and Interest Similarity The Case of CiteULike HTrsquo10]
Social Networks amp Interest Similarity
bull unilaterally connected pairs have more common itemsmetadatatags than non-connected pairsbull highest similarity for direct connections - decreasing with the increase of distance between users in SN bull reciprocal relationship users - significantly larger similarity than users in a unidirectional relationship bull traditional item-level similarity may be less reliable to find similar users in social bookmarking systemsbull peers connected by self-defined social connections could be a useful source for cross-recommendation
Conclusions
Social Web 2016 Lora Aroyo and Davide Ceolin
bull Input characteristics of items amp interests of a user into characteristics of items =gt Recommend items that feature characteristics which meet the userrsquos interests
bull Techniquesbull Data mining methods Cluster items based on their
characteristics =gt Infer usersrsquo interests into clustersbull IR methods Represent items amp users as term vectors
=gt Compute similarity between user profile vector and items
bull Utility-based methods Utility function that gets an item as input the parameters of the utility function are customized via preferences of a user
Content-based Recommendations
Social Web 2016 Lora Aroyo and Davide Ceolin
based on slides from Fabien Abel
Content Features
based on slides from Fabien Abel
User Model
based on slides from Fabien Abel
Recommendations
RecSys Issuesbull Cold-start problem (new user problem) nolittle data available to infer preferences of new
usersbull Changing User Preferences user interests may change over timebull Sparsity problem (new item problem) item descriptions are sparse eg not many user
rated or tagged an itembull Lack of Diversity (overfitting) when adapting too strongly to the preferences of users they
might see samesimilar recommendationsbull Use the right context users do things which might not be relevant for their user model
eg try out things do stuff for other peoplebull Research challenge right balance between serendipity amp personalizationbull Research challenge right way to use the influence of recommendations on userrsquos behavior
Social Web 2016 Lora Aroyo and Davide Ceolin
Announcementsbull Next deadlinesbull Tuesday March 1st 2359
Assignment 2bull Friday March 4th 1000 Post
questionbull Friday March 4th 1700 Vote
questions
image source httpwwwflickrcomphotosbionicteaching1375254387
Hands-on Teaser
bull Your Facebook Friendsrsquo popularity in a spread sheet
bull Locations of your Facebook Friendsbull Tag Cloud of your wall posts
- Social Web 2016
- Slide 2
- Slide 3
- Slide 4
- Slide 5
- Slide 6
- Slide 7
- Slide 8
- Slide 9
- Slide 10
- Slide 11
- Slide 12
- Slide 13
- Slide 14
- Slide 15
- Slide 16
- Slide 17
- Slide 18
- Slide 19
- Slide 20
- Slide 21
- Slide 22
- Slide 23
- Slide 24
- Slide 25
- Slide 26
- Slide 27
- Slide 28
- Slide 29
- Slide 30
- Slide 31
- Slide 32
- Slide 33
- Slide 34
- Slide 35
- Slide 36
- Slide 37
- Slide 38
- Slide 39
- Slide 40
- Slide 41
- Slide 42
- Slide 43
- Slide 44
- Slide 45
- Slide 46
- Slide 47
- Slide 48
- Announcements
- Slide 50
-
bull People leave traces on the Web and on their computersbull Usage data eg query logs click-through-data bull Social data eg tags (micro-)blog posts comments
bookmarks friend connections bull Documents eg pictures videosbull Personal data eg affiliations locations bull Products applications services - bought used installed
bull Not only a userrsquos behavior but also interactions of other users bull ldquopeople can make statements about merdquobull ldquopeople who are similar to me can reveal information
about merdquobull ldquosocial learningrdquo collaborative recommender systems
User amp Usage Datais Everywhere
Social Web 2016 Lora Aroyo and Davide Ceolin
bull User Profile = data structure = a characterization of a user at a particular moment represents what from a given (system) perspective there is to know about a user The data in the profile can be explicitly given by user or derived by system
bull User Model = definitions amp rules for the interpretation of observations about the user amp about the translation of that interpretation into the characteristics in a user profile user model is the recipe for obtaining amp interpreting user profiles
bull User Modeling = the process of representing the user
UM Basic Concepts
Social Web 2016 Lora Aroyo and Davide Ceolin
bull Overlay User Modeling describe user characteristics eg ldquoknowledge of a userrdquo ldquointerests of a userrdquo with respect to ldquoidealrdquo characteristics
bull Customizing user explicitly provides amp adjusts elements of the user profile
bull User model elicitation ask amp observe the user learn amp improve user profile successively ldquointeractive user modelingrdquo
bull Stereotyping stereotypical characteristics to describe a user
bull User Relevance Modeling learninfer probabilities that a given item or concept is relevant for a user
Related scientific conference httpumap2011org Related journal httpumuaiorg
User Modeling Approaches
Social Web 2016 Lora Aroyo and Davide Ceolin
httpfarm7staticflickrcom62406346803873_e756dd9bae_bjpg
Which approach suits best the conditions of
applications
bull among the oldest user modelsbull used for modeling student
knowledgebull the user is typically
characterized in terms of domain concepts amp hypotheses of the userrsquos knowledge about these concepts in relation to an (ideal) expertrsquos knowledge
bull concept-value pairs
Overlay User Models
Social Web 2016 Lora Aroyo and Davide Ceolin
bull Ask the user explicitly learnbull NLP intelligent dialoguesbull Bayesian networks Hidden Markov models
bull Observe the user learn bull Logs machine learningbull Clustering classification data mining
bull Interactive user modeling mixture of direct inputs of a user observations and inferences
User Model Elicitation
Social Web 2016 Lora Aroyo and Davide Ceolin
httphunchcomSocial Web 2016 Lora Aroyo and Davide Ceolin
Social Web 2016 Lora Aroyo and Davide Ceolin
httpfarm1staticflickrcom155413650229_31ef379b0b_bjpg
bull set of characteristics (eg attribute-value pairs) that describe a group of users
bull user is not assigned to a single stereotype - user profile can feature characteristics of several different stereotypes
User Stereotypes
based on slides from Fabien Abel
based on slides from Fabien Abel
Can we infer a Twitter-based User
Profile
User Modeling Building Blocks
based on slides from Fabien Abel
User Modeling Building Blocks
based on slides from Fabien Abel
User Modeling Building Blocks
based on slides from Fabien Abel
User Modeling Building Blocks
based on slides from Fabien Abel
Observationsbull Profile characteristics
bull Semantic enrichment solves sparsity problemsbull Profiles change over time recent profiles reflect
better current user demandsbull Temporal patterns weekend profiles differ
significantly from weekday profilesbull Impact on recommendations
bull The more fine-grained the concepts the better the recommendation performance entity-based gt topic-based gt hashtag-based
bull Semantic enrichment improves recommendation quality
bull Time-sensitivity (adapting to trends) improves performance
Social Web 2016 Lora Aroyo and Davide Ceolin
User Modelingit is not about putting everything in a user profile
it is about making the right choices
Social Web 2016 Lora Aroyo and Davide Ceolin
User Adaptation
Knowing the user to adapt a system or interfaceto improve the system functionality and user
experience
Social Web 2016 Lora Aroyo and Davide Ceolin
A Jameson Adaptive interfaces and agents The HCI handbook fundamentals evolving technologies and emerging applications pp 305ndash330 2003
User-Adaptive Systems
Social Web 2016 Lora Aroyo and Davide Ceolin
based on slides from Fabien Abel
Lastfm adapts to your music taste
bull Overfitting ldquobubble effectsrdquo loss of serendipity problem bull systems may adapt too strongly to the
interestsbehaviorbull eg an adaptive radio station may always play the
same or very similar songsbull We search for the right balance between novelty and
relevance for the userbull ldquoLost in Hyperspacerdquo problem
bull when adapting the navigation ndash ie the links on which users can click to findaccess information
bull eg re-orderinghiding of menu items may lead to confusion
Issues in User-Adaptive Systems
Social Web 2016 Lora Aroyo and Davide Ceolin
What is good user modelling amp
personalisation
httpwwwflickrcomphotosbellarosebyliz4729613108
From the consumer perspective of an adaptive system
From the provider perspective of an adaptive system
Success Perspectives
Social Web 2016 Lora Aroyo and Davide Ceolin
bull User studies askobserve (selected) people whether you did a good job
bull Log analysis Analyze (click) data and infer whether you did a good job
bull Evaluation of user modelingbull measure quality of profiles directly eg measure
overlap with existing (true) profiles or let people judge the quality of the generated user profiles
bull measure quality of application that exploits the user profile eg apply user modeling strategies in a recommender system
Evaluation Strategies
Social Web 2016 Lora Aroyo and Davide Ceolin
Evaluating User Modeling in RecSys
Social Web 2016 Lora Aroyo and Davide Ceolin
Possible Metrics
Social Web 2016 Lora Aroyo and Davide Ceolin
bull The usual IR metricsbull Precision fraction of retrieved items that are relevantbull Recall fraction of relevant items that have been
retrievedbull F-Measure (harmonic) mean of precision and recall
bull Metrics for evaluating recommendation (rankings)bull Mean Reciprocal Rank (MRR) of first relevant itembull Successk probability that relevant item occurs
within the top kbull If a true ranking is given rank correlations bull Precisionk Recallk amp F-Measurek
bull Metrics for evaluating prediction of user preferencesbull MAE = Mean Absolute Errorbull TrueFalse PositivesNegatives
Predict relevantusefulinteresting items for a given user (in a given context) itrsquos often a ranking task
Recommendation Systems
Social Web 2016 Lora Aroyo and Davide Ceolin
March 28 2013
Social Web 2016 Lora Aroyo and Davide Ceolin
Social Web 2016 Lora Aroyo and Davide Ceolin
Social Web 2016 Lora Aroyo and Davide Ceolin
httpwwwwiredcommagazine201111mf_artsyall1
Social Web 2016 Lora Aroyo and Davide Ceolin
Collaborative Filteringbull Memory-based User-Item matrix ratingspreferences of users =gt
compute similarity between users amp recommend items of similar users
bull Model-based Item-Item matrix similarity (eg based on user ratings) between items =gt recommend items that are similar to the ones the user likes
bull Model-based Clustering cluster users according to their preferences =gt recommend items of users that belong to the same cluster
bull Model-based Bayesian networks P(u likes item B | u likes item A) = how likely is it that a user who likes item A will like item B learn probabilities from user ratingspreferences
bull Others rule-based other data mining techniques
Social Web 2016 Lora Aroyo and Davide Ceolin
bull complete input data is required
bull pre-computation not possible
bull does not scale well bull high quality of
recommendations
bull abstraction (model) of input data
bull pre-computation (partially) possible (model has to be re-built from time to time)
bull scales betterbull abstraction may reduce
recommendation quality
Memory vs Model-based
Social Web 2016 Lora Aroyo and Davide Ceolin
bull collaborative filtering lsquoneighborhoodsrsquo of people with similar interest amp recommending items based on likings in neighborhood
bull limitations next to lsquocold startrsquo and lsquosparsityrsquo the lack of control (over onersquos neighborhood) is also a problem ie cannot add lsquotrustedrsquo people nor exclude lsquostrangersquo ones
bull therefore interest in lsquosocial recommendersrsquo where presence of social connections defines the similarity in interests (eg social tagging CiteULike)bull does a social connection indicate user interest similaritybull how much users interest similarity depends on the strength
of their connectionbull is it feasible to use a social network as a personalized
recommendation[Lin amp Brusilovsky Social Networks and Interest Similarity The Case of CiteULike HTrsquo10]
Social Networks amp Interest Similarity
bull unilaterally connected pairs have more common itemsmetadatatags than non-connected pairsbull highest similarity for direct connections - decreasing with the increase of distance between users in SN bull reciprocal relationship users - significantly larger similarity than users in a unidirectional relationship bull traditional item-level similarity may be less reliable to find similar users in social bookmarking systemsbull peers connected by self-defined social connections could be a useful source for cross-recommendation
Conclusions
Social Web 2016 Lora Aroyo and Davide Ceolin
bull Input characteristics of items amp interests of a user into characteristics of items =gt Recommend items that feature characteristics which meet the userrsquos interests
bull Techniquesbull Data mining methods Cluster items based on their
characteristics =gt Infer usersrsquo interests into clustersbull IR methods Represent items amp users as term vectors
=gt Compute similarity between user profile vector and items
bull Utility-based methods Utility function that gets an item as input the parameters of the utility function are customized via preferences of a user
Content-based Recommendations
Social Web 2016 Lora Aroyo and Davide Ceolin
based on slides from Fabien Abel
Content Features
based on slides from Fabien Abel
User Model
based on slides from Fabien Abel
Recommendations
RecSys Issuesbull Cold-start problem (new user problem) nolittle data available to infer preferences of new
usersbull Changing User Preferences user interests may change over timebull Sparsity problem (new item problem) item descriptions are sparse eg not many user
rated or tagged an itembull Lack of Diversity (overfitting) when adapting too strongly to the preferences of users they
might see samesimilar recommendationsbull Use the right context users do things which might not be relevant for their user model
eg try out things do stuff for other peoplebull Research challenge right balance between serendipity amp personalizationbull Research challenge right way to use the influence of recommendations on userrsquos behavior
Social Web 2016 Lora Aroyo and Davide Ceolin
Announcementsbull Next deadlinesbull Tuesday March 1st 2359
Assignment 2bull Friday March 4th 1000 Post
questionbull Friday March 4th 1700 Vote
questions
image source httpwwwflickrcomphotosbionicteaching1375254387
Hands-on Teaser
bull Your Facebook Friendsrsquo popularity in a spread sheet
bull Locations of your Facebook Friendsbull Tag Cloud of your wall posts
- Social Web 2016
- Slide 2
- Slide 3
- Slide 4
- Slide 5
- Slide 6
- Slide 7
- Slide 8
- Slide 9
- Slide 10
- Slide 11
- Slide 12
- Slide 13
- Slide 14
- Slide 15
- Slide 16
- Slide 17
- Slide 18
- Slide 19
- Slide 20
- Slide 21
- Slide 22
- Slide 23
- Slide 24
- Slide 25
- Slide 26
- Slide 27
- Slide 28
- Slide 29
- Slide 30
- Slide 31
- Slide 32
- Slide 33
- Slide 34
- Slide 35
- Slide 36
- Slide 37
- Slide 38
- Slide 39
- Slide 40
- Slide 41
- Slide 42
- Slide 43
- Slide 44
- Slide 45
- Slide 46
- Slide 47
- Slide 48
- Announcements
- Slide 50
-
bull User Profile = data structure = a characterization of a user at a particular moment represents what from a given (system) perspective there is to know about a user The data in the profile can be explicitly given by user or derived by system
bull User Model = definitions amp rules for the interpretation of observations about the user amp about the translation of that interpretation into the characteristics in a user profile user model is the recipe for obtaining amp interpreting user profiles
bull User Modeling = the process of representing the user
UM Basic Concepts
Social Web 2016 Lora Aroyo and Davide Ceolin
bull Overlay User Modeling describe user characteristics eg ldquoknowledge of a userrdquo ldquointerests of a userrdquo with respect to ldquoidealrdquo characteristics
bull Customizing user explicitly provides amp adjusts elements of the user profile
bull User model elicitation ask amp observe the user learn amp improve user profile successively ldquointeractive user modelingrdquo
bull Stereotyping stereotypical characteristics to describe a user
bull User Relevance Modeling learninfer probabilities that a given item or concept is relevant for a user
Related scientific conference httpumap2011org Related journal httpumuaiorg
User Modeling Approaches
Social Web 2016 Lora Aroyo and Davide Ceolin
httpfarm7staticflickrcom62406346803873_e756dd9bae_bjpg
Which approach suits best the conditions of
applications
bull among the oldest user modelsbull used for modeling student
knowledgebull the user is typically
characterized in terms of domain concepts amp hypotheses of the userrsquos knowledge about these concepts in relation to an (ideal) expertrsquos knowledge
bull concept-value pairs
Overlay User Models
Social Web 2016 Lora Aroyo and Davide Ceolin
bull Ask the user explicitly learnbull NLP intelligent dialoguesbull Bayesian networks Hidden Markov models
bull Observe the user learn bull Logs machine learningbull Clustering classification data mining
bull Interactive user modeling mixture of direct inputs of a user observations and inferences
User Model Elicitation
Social Web 2016 Lora Aroyo and Davide Ceolin
httphunchcomSocial Web 2016 Lora Aroyo and Davide Ceolin
Social Web 2016 Lora Aroyo and Davide Ceolin
httpfarm1staticflickrcom155413650229_31ef379b0b_bjpg
bull set of characteristics (eg attribute-value pairs) that describe a group of users
bull user is not assigned to a single stereotype - user profile can feature characteristics of several different stereotypes
User Stereotypes
based on slides from Fabien Abel
based on slides from Fabien Abel
Can we infer a Twitter-based User
Profile
User Modeling Building Blocks
based on slides from Fabien Abel
User Modeling Building Blocks
based on slides from Fabien Abel
User Modeling Building Blocks
based on slides from Fabien Abel
User Modeling Building Blocks
based on slides from Fabien Abel
Observationsbull Profile characteristics
bull Semantic enrichment solves sparsity problemsbull Profiles change over time recent profiles reflect
better current user demandsbull Temporal patterns weekend profiles differ
significantly from weekday profilesbull Impact on recommendations
bull The more fine-grained the concepts the better the recommendation performance entity-based gt topic-based gt hashtag-based
bull Semantic enrichment improves recommendation quality
bull Time-sensitivity (adapting to trends) improves performance
Social Web 2016 Lora Aroyo and Davide Ceolin
User Modelingit is not about putting everything in a user profile
it is about making the right choices
Social Web 2016 Lora Aroyo and Davide Ceolin
User Adaptation
Knowing the user to adapt a system or interfaceto improve the system functionality and user
experience
Social Web 2016 Lora Aroyo and Davide Ceolin
A Jameson Adaptive interfaces and agents The HCI handbook fundamentals evolving technologies and emerging applications pp 305ndash330 2003
User-Adaptive Systems
Social Web 2016 Lora Aroyo and Davide Ceolin
based on slides from Fabien Abel
Lastfm adapts to your music taste
bull Overfitting ldquobubble effectsrdquo loss of serendipity problem bull systems may adapt too strongly to the
interestsbehaviorbull eg an adaptive radio station may always play the
same or very similar songsbull We search for the right balance between novelty and
relevance for the userbull ldquoLost in Hyperspacerdquo problem
bull when adapting the navigation ndash ie the links on which users can click to findaccess information
bull eg re-orderinghiding of menu items may lead to confusion
Issues in User-Adaptive Systems
Social Web 2016 Lora Aroyo and Davide Ceolin
What is good user modelling amp
personalisation
httpwwwflickrcomphotosbellarosebyliz4729613108
From the consumer perspective of an adaptive system
From the provider perspective of an adaptive system
Success Perspectives
Social Web 2016 Lora Aroyo and Davide Ceolin
bull User studies askobserve (selected) people whether you did a good job
bull Log analysis Analyze (click) data and infer whether you did a good job
bull Evaluation of user modelingbull measure quality of profiles directly eg measure
overlap with existing (true) profiles or let people judge the quality of the generated user profiles
bull measure quality of application that exploits the user profile eg apply user modeling strategies in a recommender system
Evaluation Strategies
Social Web 2016 Lora Aroyo and Davide Ceolin
Evaluating User Modeling in RecSys
Social Web 2016 Lora Aroyo and Davide Ceolin
Possible Metrics
Social Web 2016 Lora Aroyo and Davide Ceolin
bull The usual IR metricsbull Precision fraction of retrieved items that are relevantbull Recall fraction of relevant items that have been
retrievedbull F-Measure (harmonic) mean of precision and recall
bull Metrics for evaluating recommendation (rankings)bull Mean Reciprocal Rank (MRR) of first relevant itembull Successk probability that relevant item occurs
within the top kbull If a true ranking is given rank correlations bull Precisionk Recallk amp F-Measurek
bull Metrics for evaluating prediction of user preferencesbull MAE = Mean Absolute Errorbull TrueFalse PositivesNegatives
Predict relevantusefulinteresting items for a given user (in a given context) itrsquos often a ranking task
Recommendation Systems
Social Web 2016 Lora Aroyo and Davide Ceolin
March 28 2013
Social Web 2016 Lora Aroyo and Davide Ceolin
Social Web 2016 Lora Aroyo and Davide Ceolin
Social Web 2016 Lora Aroyo and Davide Ceolin
httpwwwwiredcommagazine201111mf_artsyall1
Social Web 2016 Lora Aroyo and Davide Ceolin
Collaborative Filteringbull Memory-based User-Item matrix ratingspreferences of users =gt
compute similarity between users amp recommend items of similar users
bull Model-based Item-Item matrix similarity (eg based on user ratings) between items =gt recommend items that are similar to the ones the user likes
bull Model-based Clustering cluster users according to their preferences =gt recommend items of users that belong to the same cluster
bull Model-based Bayesian networks P(u likes item B | u likes item A) = how likely is it that a user who likes item A will like item B learn probabilities from user ratingspreferences
bull Others rule-based other data mining techniques
Social Web 2016 Lora Aroyo and Davide Ceolin
bull complete input data is required
bull pre-computation not possible
bull does not scale well bull high quality of
recommendations
bull abstraction (model) of input data
bull pre-computation (partially) possible (model has to be re-built from time to time)
bull scales betterbull abstraction may reduce
recommendation quality
Memory vs Model-based
Social Web 2016 Lora Aroyo and Davide Ceolin
bull collaborative filtering lsquoneighborhoodsrsquo of people with similar interest amp recommending items based on likings in neighborhood
bull limitations next to lsquocold startrsquo and lsquosparsityrsquo the lack of control (over onersquos neighborhood) is also a problem ie cannot add lsquotrustedrsquo people nor exclude lsquostrangersquo ones
bull therefore interest in lsquosocial recommendersrsquo where presence of social connections defines the similarity in interests (eg social tagging CiteULike)bull does a social connection indicate user interest similaritybull how much users interest similarity depends on the strength
of their connectionbull is it feasible to use a social network as a personalized
recommendation[Lin amp Brusilovsky Social Networks and Interest Similarity The Case of CiteULike HTrsquo10]
Social Networks amp Interest Similarity
bull unilaterally connected pairs have more common itemsmetadatatags than non-connected pairsbull highest similarity for direct connections - decreasing with the increase of distance between users in SN bull reciprocal relationship users - significantly larger similarity than users in a unidirectional relationship bull traditional item-level similarity may be less reliable to find similar users in social bookmarking systemsbull peers connected by self-defined social connections could be a useful source for cross-recommendation
Conclusions
Social Web 2016 Lora Aroyo and Davide Ceolin
bull Input characteristics of items amp interests of a user into characteristics of items =gt Recommend items that feature characteristics which meet the userrsquos interests
bull Techniquesbull Data mining methods Cluster items based on their
characteristics =gt Infer usersrsquo interests into clustersbull IR methods Represent items amp users as term vectors
=gt Compute similarity between user profile vector and items
bull Utility-based methods Utility function that gets an item as input the parameters of the utility function are customized via preferences of a user
Content-based Recommendations
Social Web 2016 Lora Aroyo and Davide Ceolin
based on slides from Fabien Abel
Content Features
based on slides from Fabien Abel
User Model
based on slides from Fabien Abel
Recommendations
RecSys Issuesbull Cold-start problem (new user problem) nolittle data available to infer preferences of new
usersbull Changing User Preferences user interests may change over timebull Sparsity problem (new item problem) item descriptions are sparse eg not many user
rated or tagged an itembull Lack of Diversity (overfitting) when adapting too strongly to the preferences of users they
might see samesimilar recommendationsbull Use the right context users do things which might not be relevant for their user model
eg try out things do stuff for other peoplebull Research challenge right balance between serendipity amp personalizationbull Research challenge right way to use the influence of recommendations on userrsquos behavior
Social Web 2016 Lora Aroyo and Davide Ceolin
Announcementsbull Next deadlinesbull Tuesday March 1st 2359
Assignment 2bull Friday March 4th 1000 Post
questionbull Friday March 4th 1700 Vote
questions
image source httpwwwflickrcomphotosbionicteaching1375254387
Hands-on Teaser
bull Your Facebook Friendsrsquo popularity in a spread sheet
bull Locations of your Facebook Friendsbull Tag Cloud of your wall posts
- Social Web 2016
- Slide 2
- Slide 3
- Slide 4
- Slide 5
- Slide 6
- Slide 7
- Slide 8
- Slide 9
- Slide 10
- Slide 11
- Slide 12
- Slide 13
- Slide 14
- Slide 15
- Slide 16
- Slide 17
- Slide 18
- Slide 19
- Slide 20
- Slide 21
- Slide 22
- Slide 23
- Slide 24
- Slide 25
- Slide 26
- Slide 27
- Slide 28
- Slide 29
- Slide 30
- Slide 31
- Slide 32
- Slide 33
- Slide 34
- Slide 35
- Slide 36
- Slide 37
- Slide 38
- Slide 39
- Slide 40
- Slide 41
- Slide 42
- Slide 43
- Slide 44
- Slide 45
- Slide 46
- Slide 47
- Slide 48
- Announcements
- Slide 50
-
bull Overlay User Modeling describe user characteristics eg ldquoknowledge of a userrdquo ldquointerests of a userrdquo with respect to ldquoidealrdquo characteristics
bull Customizing user explicitly provides amp adjusts elements of the user profile
bull User model elicitation ask amp observe the user learn amp improve user profile successively ldquointeractive user modelingrdquo
bull Stereotyping stereotypical characteristics to describe a user
bull User Relevance Modeling learninfer probabilities that a given item or concept is relevant for a user
Related scientific conference httpumap2011org Related journal httpumuaiorg
User Modeling Approaches
Social Web 2016 Lora Aroyo and Davide Ceolin
httpfarm7staticflickrcom62406346803873_e756dd9bae_bjpg
Which approach suits best the conditions of
applications
bull among the oldest user modelsbull used for modeling student
knowledgebull the user is typically
characterized in terms of domain concepts amp hypotheses of the userrsquos knowledge about these concepts in relation to an (ideal) expertrsquos knowledge
bull concept-value pairs
Overlay User Models
Social Web 2016 Lora Aroyo and Davide Ceolin
bull Ask the user explicitly learnbull NLP intelligent dialoguesbull Bayesian networks Hidden Markov models
bull Observe the user learn bull Logs machine learningbull Clustering classification data mining
bull Interactive user modeling mixture of direct inputs of a user observations and inferences
User Model Elicitation
Social Web 2016 Lora Aroyo and Davide Ceolin
httphunchcomSocial Web 2016 Lora Aroyo and Davide Ceolin
Social Web 2016 Lora Aroyo and Davide Ceolin
httpfarm1staticflickrcom155413650229_31ef379b0b_bjpg
bull set of characteristics (eg attribute-value pairs) that describe a group of users
bull user is not assigned to a single stereotype - user profile can feature characteristics of several different stereotypes
User Stereotypes
based on slides from Fabien Abel
based on slides from Fabien Abel
Can we infer a Twitter-based User
Profile
User Modeling Building Blocks
based on slides from Fabien Abel
User Modeling Building Blocks
based on slides from Fabien Abel
User Modeling Building Blocks
based on slides from Fabien Abel
User Modeling Building Blocks
based on slides from Fabien Abel
Observationsbull Profile characteristics
bull Semantic enrichment solves sparsity problemsbull Profiles change over time recent profiles reflect
better current user demandsbull Temporal patterns weekend profiles differ
significantly from weekday profilesbull Impact on recommendations
bull The more fine-grained the concepts the better the recommendation performance entity-based gt topic-based gt hashtag-based
bull Semantic enrichment improves recommendation quality
bull Time-sensitivity (adapting to trends) improves performance
Social Web 2016 Lora Aroyo and Davide Ceolin
User Modelingit is not about putting everything in a user profile
it is about making the right choices
Social Web 2016 Lora Aroyo and Davide Ceolin
User Adaptation
Knowing the user to adapt a system or interfaceto improve the system functionality and user
experience
Social Web 2016 Lora Aroyo and Davide Ceolin
A Jameson Adaptive interfaces and agents The HCI handbook fundamentals evolving technologies and emerging applications pp 305ndash330 2003
User-Adaptive Systems
Social Web 2016 Lora Aroyo and Davide Ceolin
based on slides from Fabien Abel
Lastfm adapts to your music taste
bull Overfitting ldquobubble effectsrdquo loss of serendipity problem bull systems may adapt too strongly to the
interestsbehaviorbull eg an adaptive radio station may always play the
same or very similar songsbull We search for the right balance between novelty and
relevance for the userbull ldquoLost in Hyperspacerdquo problem
bull when adapting the navigation ndash ie the links on which users can click to findaccess information
bull eg re-orderinghiding of menu items may lead to confusion
Issues in User-Adaptive Systems
Social Web 2016 Lora Aroyo and Davide Ceolin
What is good user modelling amp
personalisation
httpwwwflickrcomphotosbellarosebyliz4729613108
From the consumer perspective of an adaptive system
From the provider perspective of an adaptive system
Success Perspectives
Social Web 2016 Lora Aroyo and Davide Ceolin
bull User studies askobserve (selected) people whether you did a good job
bull Log analysis Analyze (click) data and infer whether you did a good job
bull Evaluation of user modelingbull measure quality of profiles directly eg measure
overlap with existing (true) profiles or let people judge the quality of the generated user profiles
bull measure quality of application that exploits the user profile eg apply user modeling strategies in a recommender system
Evaluation Strategies
Social Web 2016 Lora Aroyo and Davide Ceolin
Evaluating User Modeling in RecSys
Social Web 2016 Lora Aroyo and Davide Ceolin
Possible Metrics
Social Web 2016 Lora Aroyo and Davide Ceolin
bull The usual IR metricsbull Precision fraction of retrieved items that are relevantbull Recall fraction of relevant items that have been
retrievedbull F-Measure (harmonic) mean of precision and recall
bull Metrics for evaluating recommendation (rankings)bull Mean Reciprocal Rank (MRR) of first relevant itembull Successk probability that relevant item occurs
within the top kbull If a true ranking is given rank correlations bull Precisionk Recallk amp F-Measurek
bull Metrics for evaluating prediction of user preferencesbull MAE = Mean Absolute Errorbull TrueFalse PositivesNegatives
Predict relevantusefulinteresting items for a given user (in a given context) itrsquos often a ranking task
Recommendation Systems
Social Web 2016 Lora Aroyo and Davide Ceolin
March 28 2013
Social Web 2016 Lora Aroyo and Davide Ceolin
Social Web 2016 Lora Aroyo and Davide Ceolin
Social Web 2016 Lora Aroyo and Davide Ceolin
httpwwwwiredcommagazine201111mf_artsyall1
Social Web 2016 Lora Aroyo and Davide Ceolin
Collaborative Filteringbull Memory-based User-Item matrix ratingspreferences of users =gt
compute similarity between users amp recommend items of similar users
bull Model-based Item-Item matrix similarity (eg based on user ratings) between items =gt recommend items that are similar to the ones the user likes
bull Model-based Clustering cluster users according to their preferences =gt recommend items of users that belong to the same cluster
bull Model-based Bayesian networks P(u likes item B | u likes item A) = how likely is it that a user who likes item A will like item B learn probabilities from user ratingspreferences
bull Others rule-based other data mining techniques
Social Web 2016 Lora Aroyo and Davide Ceolin
bull complete input data is required
bull pre-computation not possible
bull does not scale well bull high quality of
recommendations
bull abstraction (model) of input data
bull pre-computation (partially) possible (model has to be re-built from time to time)
bull scales betterbull abstraction may reduce
recommendation quality
Memory vs Model-based
Social Web 2016 Lora Aroyo and Davide Ceolin
bull collaborative filtering lsquoneighborhoodsrsquo of people with similar interest amp recommending items based on likings in neighborhood
bull limitations next to lsquocold startrsquo and lsquosparsityrsquo the lack of control (over onersquos neighborhood) is also a problem ie cannot add lsquotrustedrsquo people nor exclude lsquostrangersquo ones
bull therefore interest in lsquosocial recommendersrsquo where presence of social connections defines the similarity in interests (eg social tagging CiteULike)bull does a social connection indicate user interest similaritybull how much users interest similarity depends on the strength
of their connectionbull is it feasible to use a social network as a personalized
recommendation[Lin amp Brusilovsky Social Networks and Interest Similarity The Case of CiteULike HTrsquo10]
Social Networks amp Interest Similarity
bull unilaterally connected pairs have more common itemsmetadatatags than non-connected pairsbull highest similarity for direct connections - decreasing with the increase of distance between users in SN bull reciprocal relationship users - significantly larger similarity than users in a unidirectional relationship bull traditional item-level similarity may be less reliable to find similar users in social bookmarking systemsbull peers connected by self-defined social connections could be a useful source for cross-recommendation
Conclusions
Social Web 2016 Lora Aroyo and Davide Ceolin
bull Input characteristics of items amp interests of a user into characteristics of items =gt Recommend items that feature characteristics which meet the userrsquos interests
bull Techniquesbull Data mining methods Cluster items based on their
characteristics =gt Infer usersrsquo interests into clustersbull IR methods Represent items amp users as term vectors
=gt Compute similarity between user profile vector and items
bull Utility-based methods Utility function that gets an item as input the parameters of the utility function are customized via preferences of a user
Content-based Recommendations
Social Web 2016 Lora Aroyo and Davide Ceolin
based on slides from Fabien Abel
Content Features
based on slides from Fabien Abel
User Model
based on slides from Fabien Abel
Recommendations
RecSys Issuesbull Cold-start problem (new user problem) nolittle data available to infer preferences of new
usersbull Changing User Preferences user interests may change over timebull Sparsity problem (new item problem) item descriptions are sparse eg not many user
rated or tagged an itembull Lack of Diversity (overfitting) when adapting too strongly to the preferences of users they
might see samesimilar recommendationsbull Use the right context users do things which might not be relevant for their user model
eg try out things do stuff for other peoplebull Research challenge right balance between serendipity amp personalizationbull Research challenge right way to use the influence of recommendations on userrsquos behavior
Social Web 2016 Lora Aroyo and Davide Ceolin
Announcementsbull Next deadlinesbull Tuesday March 1st 2359
Assignment 2bull Friday March 4th 1000 Post
questionbull Friday March 4th 1700 Vote
questions
image source httpwwwflickrcomphotosbionicteaching1375254387
Hands-on Teaser
bull Your Facebook Friendsrsquo popularity in a spread sheet
bull Locations of your Facebook Friendsbull Tag Cloud of your wall posts
- Social Web 2016
- Slide 2
- Slide 3
- Slide 4
- Slide 5
- Slide 6
- Slide 7
- Slide 8
- Slide 9
- Slide 10
- Slide 11
- Slide 12
- Slide 13
- Slide 14
- Slide 15
- Slide 16
- Slide 17
- Slide 18
- Slide 19
- Slide 20
- Slide 21
- Slide 22
- Slide 23
- Slide 24
- Slide 25
- Slide 26
- Slide 27
- Slide 28
- Slide 29
- Slide 30
- Slide 31
- Slide 32
- Slide 33
- Slide 34
- Slide 35
- Slide 36
- Slide 37
- Slide 38
- Slide 39
- Slide 40
- Slide 41
- Slide 42
- Slide 43
- Slide 44
- Slide 45
- Slide 46
- Slide 47
- Slide 48
- Announcements
- Slide 50
-
httpfarm7staticflickrcom62406346803873_e756dd9bae_bjpg
Which approach suits best the conditions of
applications
bull among the oldest user modelsbull used for modeling student
knowledgebull the user is typically
characterized in terms of domain concepts amp hypotheses of the userrsquos knowledge about these concepts in relation to an (ideal) expertrsquos knowledge
bull concept-value pairs
Overlay User Models
Social Web 2016 Lora Aroyo and Davide Ceolin
bull Ask the user explicitly learnbull NLP intelligent dialoguesbull Bayesian networks Hidden Markov models
bull Observe the user learn bull Logs machine learningbull Clustering classification data mining
bull Interactive user modeling mixture of direct inputs of a user observations and inferences
User Model Elicitation
Social Web 2016 Lora Aroyo and Davide Ceolin
httphunchcomSocial Web 2016 Lora Aroyo and Davide Ceolin
Social Web 2016 Lora Aroyo and Davide Ceolin
httpfarm1staticflickrcom155413650229_31ef379b0b_bjpg
bull set of characteristics (eg attribute-value pairs) that describe a group of users
bull user is not assigned to a single stereotype - user profile can feature characteristics of several different stereotypes
User Stereotypes
based on slides from Fabien Abel
based on slides from Fabien Abel
Can we infer a Twitter-based User
Profile
User Modeling Building Blocks
based on slides from Fabien Abel
User Modeling Building Blocks
based on slides from Fabien Abel
User Modeling Building Blocks
based on slides from Fabien Abel
User Modeling Building Blocks
based on slides from Fabien Abel
Observationsbull Profile characteristics
bull Semantic enrichment solves sparsity problemsbull Profiles change over time recent profiles reflect
better current user demandsbull Temporal patterns weekend profiles differ
significantly from weekday profilesbull Impact on recommendations
bull The more fine-grained the concepts the better the recommendation performance entity-based gt topic-based gt hashtag-based
bull Semantic enrichment improves recommendation quality
bull Time-sensitivity (adapting to trends) improves performance
Social Web 2016 Lora Aroyo and Davide Ceolin
User Modelingit is not about putting everything in a user profile
it is about making the right choices
Social Web 2016 Lora Aroyo and Davide Ceolin
User Adaptation
Knowing the user to adapt a system or interfaceto improve the system functionality and user
experience
Social Web 2016 Lora Aroyo and Davide Ceolin
A Jameson Adaptive interfaces and agents The HCI handbook fundamentals evolving technologies and emerging applications pp 305ndash330 2003
User-Adaptive Systems
Social Web 2016 Lora Aroyo and Davide Ceolin
based on slides from Fabien Abel
Lastfm adapts to your music taste
bull Overfitting ldquobubble effectsrdquo loss of serendipity problem bull systems may adapt too strongly to the
interestsbehaviorbull eg an adaptive radio station may always play the
same or very similar songsbull We search for the right balance between novelty and
relevance for the userbull ldquoLost in Hyperspacerdquo problem
bull when adapting the navigation ndash ie the links on which users can click to findaccess information
bull eg re-orderinghiding of menu items may lead to confusion
Issues in User-Adaptive Systems
Social Web 2016 Lora Aroyo and Davide Ceolin
What is good user modelling amp
personalisation
httpwwwflickrcomphotosbellarosebyliz4729613108
From the consumer perspective of an adaptive system
From the provider perspective of an adaptive system
Success Perspectives
Social Web 2016 Lora Aroyo and Davide Ceolin
bull User studies askobserve (selected) people whether you did a good job
bull Log analysis Analyze (click) data and infer whether you did a good job
bull Evaluation of user modelingbull measure quality of profiles directly eg measure
overlap with existing (true) profiles or let people judge the quality of the generated user profiles
bull measure quality of application that exploits the user profile eg apply user modeling strategies in a recommender system
Evaluation Strategies
Social Web 2016 Lora Aroyo and Davide Ceolin
Evaluating User Modeling in RecSys
Social Web 2016 Lora Aroyo and Davide Ceolin
Possible Metrics
Social Web 2016 Lora Aroyo and Davide Ceolin
bull The usual IR metricsbull Precision fraction of retrieved items that are relevantbull Recall fraction of relevant items that have been
retrievedbull F-Measure (harmonic) mean of precision and recall
bull Metrics for evaluating recommendation (rankings)bull Mean Reciprocal Rank (MRR) of first relevant itembull Successk probability that relevant item occurs
within the top kbull If a true ranking is given rank correlations bull Precisionk Recallk amp F-Measurek
bull Metrics for evaluating prediction of user preferencesbull MAE = Mean Absolute Errorbull TrueFalse PositivesNegatives
Predict relevantusefulinteresting items for a given user (in a given context) itrsquos often a ranking task
Recommendation Systems
Social Web 2016 Lora Aroyo and Davide Ceolin
March 28 2013
Social Web 2016 Lora Aroyo and Davide Ceolin
Social Web 2016 Lora Aroyo and Davide Ceolin
Social Web 2016 Lora Aroyo and Davide Ceolin
httpwwwwiredcommagazine201111mf_artsyall1
Social Web 2016 Lora Aroyo and Davide Ceolin
Collaborative Filteringbull Memory-based User-Item matrix ratingspreferences of users =gt
compute similarity between users amp recommend items of similar users
bull Model-based Item-Item matrix similarity (eg based on user ratings) between items =gt recommend items that are similar to the ones the user likes
bull Model-based Clustering cluster users according to their preferences =gt recommend items of users that belong to the same cluster
bull Model-based Bayesian networks P(u likes item B | u likes item A) = how likely is it that a user who likes item A will like item B learn probabilities from user ratingspreferences
bull Others rule-based other data mining techniques
Social Web 2016 Lora Aroyo and Davide Ceolin
bull complete input data is required
bull pre-computation not possible
bull does not scale well bull high quality of
recommendations
bull abstraction (model) of input data
bull pre-computation (partially) possible (model has to be re-built from time to time)
bull scales betterbull abstraction may reduce
recommendation quality
Memory vs Model-based
Social Web 2016 Lora Aroyo and Davide Ceolin
bull collaborative filtering lsquoneighborhoodsrsquo of people with similar interest amp recommending items based on likings in neighborhood
bull limitations next to lsquocold startrsquo and lsquosparsityrsquo the lack of control (over onersquos neighborhood) is also a problem ie cannot add lsquotrustedrsquo people nor exclude lsquostrangersquo ones
bull therefore interest in lsquosocial recommendersrsquo where presence of social connections defines the similarity in interests (eg social tagging CiteULike)bull does a social connection indicate user interest similaritybull how much users interest similarity depends on the strength
of their connectionbull is it feasible to use a social network as a personalized
recommendation[Lin amp Brusilovsky Social Networks and Interest Similarity The Case of CiteULike HTrsquo10]
Social Networks amp Interest Similarity
bull unilaterally connected pairs have more common itemsmetadatatags than non-connected pairsbull highest similarity for direct connections - decreasing with the increase of distance between users in SN bull reciprocal relationship users - significantly larger similarity than users in a unidirectional relationship bull traditional item-level similarity may be less reliable to find similar users in social bookmarking systemsbull peers connected by self-defined social connections could be a useful source for cross-recommendation
Conclusions
Social Web 2016 Lora Aroyo and Davide Ceolin
bull Input characteristics of items amp interests of a user into characteristics of items =gt Recommend items that feature characteristics which meet the userrsquos interests
bull Techniquesbull Data mining methods Cluster items based on their
characteristics =gt Infer usersrsquo interests into clustersbull IR methods Represent items amp users as term vectors
=gt Compute similarity between user profile vector and items
bull Utility-based methods Utility function that gets an item as input the parameters of the utility function are customized via preferences of a user
Content-based Recommendations
Social Web 2016 Lora Aroyo and Davide Ceolin
based on slides from Fabien Abel
Content Features
based on slides from Fabien Abel
User Model
based on slides from Fabien Abel
Recommendations
RecSys Issuesbull Cold-start problem (new user problem) nolittle data available to infer preferences of new
usersbull Changing User Preferences user interests may change over timebull Sparsity problem (new item problem) item descriptions are sparse eg not many user
rated or tagged an itembull Lack of Diversity (overfitting) when adapting too strongly to the preferences of users they
might see samesimilar recommendationsbull Use the right context users do things which might not be relevant for their user model
eg try out things do stuff for other peoplebull Research challenge right balance between serendipity amp personalizationbull Research challenge right way to use the influence of recommendations on userrsquos behavior
Social Web 2016 Lora Aroyo and Davide Ceolin
Announcementsbull Next deadlinesbull Tuesday March 1st 2359
Assignment 2bull Friday March 4th 1000 Post
questionbull Friday March 4th 1700 Vote
questions
image source httpwwwflickrcomphotosbionicteaching1375254387
Hands-on Teaser
bull Your Facebook Friendsrsquo popularity in a spread sheet
bull Locations of your Facebook Friendsbull Tag Cloud of your wall posts
- Social Web 2016
- Slide 2
- Slide 3
- Slide 4
- Slide 5
- Slide 6
- Slide 7
- Slide 8
- Slide 9
- Slide 10
- Slide 11
- Slide 12
- Slide 13
- Slide 14
- Slide 15
- Slide 16
- Slide 17
- Slide 18
- Slide 19
- Slide 20
- Slide 21
- Slide 22
- Slide 23
- Slide 24
- Slide 25
- Slide 26
- Slide 27
- Slide 28
- Slide 29
- Slide 30
- Slide 31
- Slide 32
- Slide 33
- Slide 34
- Slide 35
- Slide 36
- Slide 37
- Slide 38
- Slide 39
- Slide 40
- Slide 41
- Slide 42
- Slide 43
- Slide 44
- Slide 45
- Slide 46
- Slide 47
- Slide 48
- Announcements
- Slide 50
-
bull among the oldest user modelsbull used for modeling student
knowledgebull the user is typically
characterized in terms of domain concepts amp hypotheses of the userrsquos knowledge about these concepts in relation to an (ideal) expertrsquos knowledge
bull concept-value pairs
Overlay User Models
Social Web 2016 Lora Aroyo and Davide Ceolin
bull Ask the user explicitly learnbull NLP intelligent dialoguesbull Bayesian networks Hidden Markov models
bull Observe the user learn bull Logs machine learningbull Clustering classification data mining
bull Interactive user modeling mixture of direct inputs of a user observations and inferences
User Model Elicitation
Social Web 2016 Lora Aroyo and Davide Ceolin
httphunchcomSocial Web 2016 Lora Aroyo and Davide Ceolin
Social Web 2016 Lora Aroyo and Davide Ceolin
httpfarm1staticflickrcom155413650229_31ef379b0b_bjpg
bull set of characteristics (eg attribute-value pairs) that describe a group of users
bull user is not assigned to a single stereotype - user profile can feature characteristics of several different stereotypes
User Stereotypes
based on slides from Fabien Abel
based on slides from Fabien Abel
Can we infer a Twitter-based User
Profile
User Modeling Building Blocks
based on slides from Fabien Abel
User Modeling Building Blocks
based on slides from Fabien Abel
User Modeling Building Blocks
based on slides from Fabien Abel
User Modeling Building Blocks
based on slides from Fabien Abel
Observationsbull Profile characteristics
bull Semantic enrichment solves sparsity problemsbull Profiles change over time recent profiles reflect
better current user demandsbull Temporal patterns weekend profiles differ
significantly from weekday profilesbull Impact on recommendations
bull The more fine-grained the concepts the better the recommendation performance entity-based gt topic-based gt hashtag-based
bull Semantic enrichment improves recommendation quality
bull Time-sensitivity (adapting to trends) improves performance
Social Web 2016 Lora Aroyo and Davide Ceolin
User Modelingit is not about putting everything in a user profile
it is about making the right choices
Social Web 2016 Lora Aroyo and Davide Ceolin
User Adaptation
Knowing the user to adapt a system or interfaceto improve the system functionality and user
experience
Social Web 2016 Lora Aroyo and Davide Ceolin
A Jameson Adaptive interfaces and agents The HCI handbook fundamentals evolving technologies and emerging applications pp 305ndash330 2003
User-Adaptive Systems
Social Web 2016 Lora Aroyo and Davide Ceolin
based on slides from Fabien Abel
Lastfm adapts to your music taste
bull Overfitting ldquobubble effectsrdquo loss of serendipity problem bull systems may adapt too strongly to the
interestsbehaviorbull eg an adaptive radio station may always play the
same or very similar songsbull We search for the right balance between novelty and
relevance for the userbull ldquoLost in Hyperspacerdquo problem
bull when adapting the navigation ndash ie the links on which users can click to findaccess information
bull eg re-orderinghiding of menu items may lead to confusion
Issues in User-Adaptive Systems
Social Web 2016 Lora Aroyo and Davide Ceolin
What is good user modelling amp
personalisation
httpwwwflickrcomphotosbellarosebyliz4729613108
From the consumer perspective of an adaptive system
From the provider perspective of an adaptive system
Success Perspectives
Social Web 2016 Lora Aroyo and Davide Ceolin
bull User studies askobserve (selected) people whether you did a good job
bull Log analysis Analyze (click) data and infer whether you did a good job
bull Evaluation of user modelingbull measure quality of profiles directly eg measure
overlap with existing (true) profiles or let people judge the quality of the generated user profiles
bull measure quality of application that exploits the user profile eg apply user modeling strategies in a recommender system
Evaluation Strategies
Social Web 2016 Lora Aroyo and Davide Ceolin
Evaluating User Modeling in RecSys
Social Web 2016 Lora Aroyo and Davide Ceolin
Possible Metrics
Social Web 2016 Lora Aroyo and Davide Ceolin
bull The usual IR metricsbull Precision fraction of retrieved items that are relevantbull Recall fraction of relevant items that have been
retrievedbull F-Measure (harmonic) mean of precision and recall
bull Metrics for evaluating recommendation (rankings)bull Mean Reciprocal Rank (MRR) of first relevant itembull Successk probability that relevant item occurs
within the top kbull If a true ranking is given rank correlations bull Precisionk Recallk amp F-Measurek
bull Metrics for evaluating prediction of user preferencesbull MAE = Mean Absolute Errorbull TrueFalse PositivesNegatives
Predict relevantusefulinteresting items for a given user (in a given context) itrsquos often a ranking task
Recommendation Systems
Social Web 2016 Lora Aroyo and Davide Ceolin
March 28 2013
Social Web 2016 Lora Aroyo and Davide Ceolin
Social Web 2016 Lora Aroyo and Davide Ceolin
Social Web 2016 Lora Aroyo and Davide Ceolin
httpwwwwiredcommagazine201111mf_artsyall1
Social Web 2016 Lora Aroyo and Davide Ceolin
Collaborative Filteringbull Memory-based User-Item matrix ratingspreferences of users =gt
compute similarity between users amp recommend items of similar users
bull Model-based Item-Item matrix similarity (eg based on user ratings) between items =gt recommend items that are similar to the ones the user likes
bull Model-based Clustering cluster users according to their preferences =gt recommend items of users that belong to the same cluster
bull Model-based Bayesian networks P(u likes item B | u likes item A) = how likely is it that a user who likes item A will like item B learn probabilities from user ratingspreferences
bull Others rule-based other data mining techniques
Social Web 2016 Lora Aroyo and Davide Ceolin
bull complete input data is required
bull pre-computation not possible
bull does not scale well bull high quality of
recommendations
bull abstraction (model) of input data
bull pre-computation (partially) possible (model has to be re-built from time to time)
bull scales betterbull abstraction may reduce
recommendation quality
Memory vs Model-based
Social Web 2016 Lora Aroyo and Davide Ceolin
bull collaborative filtering lsquoneighborhoodsrsquo of people with similar interest amp recommending items based on likings in neighborhood
bull limitations next to lsquocold startrsquo and lsquosparsityrsquo the lack of control (over onersquos neighborhood) is also a problem ie cannot add lsquotrustedrsquo people nor exclude lsquostrangersquo ones
bull therefore interest in lsquosocial recommendersrsquo where presence of social connections defines the similarity in interests (eg social tagging CiteULike)bull does a social connection indicate user interest similaritybull how much users interest similarity depends on the strength
of their connectionbull is it feasible to use a social network as a personalized
recommendation[Lin amp Brusilovsky Social Networks and Interest Similarity The Case of CiteULike HTrsquo10]
Social Networks amp Interest Similarity
bull unilaterally connected pairs have more common itemsmetadatatags than non-connected pairsbull highest similarity for direct connections - decreasing with the increase of distance between users in SN bull reciprocal relationship users - significantly larger similarity than users in a unidirectional relationship bull traditional item-level similarity may be less reliable to find similar users in social bookmarking systemsbull peers connected by self-defined social connections could be a useful source for cross-recommendation
Conclusions
Social Web 2016 Lora Aroyo and Davide Ceolin
bull Input characteristics of items amp interests of a user into characteristics of items =gt Recommend items that feature characteristics which meet the userrsquos interests
bull Techniquesbull Data mining methods Cluster items based on their
characteristics =gt Infer usersrsquo interests into clustersbull IR methods Represent items amp users as term vectors
=gt Compute similarity between user profile vector and items
bull Utility-based methods Utility function that gets an item as input the parameters of the utility function are customized via preferences of a user
Content-based Recommendations
Social Web 2016 Lora Aroyo and Davide Ceolin
based on slides from Fabien Abel
Content Features
based on slides from Fabien Abel
User Model
based on slides from Fabien Abel
Recommendations
RecSys Issuesbull Cold-start problem (new user problem) nolittle data available to infer preferences of new
usersbull Changing User Preferences user interests may change over timebull Sparsity problem (new item problem) item descriptions are sparse eg not many user
rated or tagged an itembull Lack of Diversity (overfitting) when adapting too strongly to the preferences of users they
might see samesimilar recommendationsbull Use the right context users do things which might not be relevant for their user model
eg try out things do stuff for other peoplebull Research challenge right balance between serendipity amp personalizationbull Research challenge right way to use the influence of recommendations on userrsquos behavior
Social Web 2016 Lora Aroyo and Davide Ceolin
Announcementsbull Next deadlinesbull Tuesday March 1st 2359
Assignment 2bull Friday March 4th 1000 Post
questionbull Friday March 4th 1700 Vote
questions
image source httpwwwflickrcomphotosbionicteaching1375254387
Hands-on Teaser
bull Your Facebook Friendsrsquo popularity in a spread sheet
bull Locations of your Facebook Friendsbull Tag Cloud of your wall posts
- Social Web 2016
- Slide 2
- Slide 3
- Slide 4
- Slide 5
- Slide 6
- Slide 7
- Slide 8
- Slide 9
- Slide 10
- Slide 11
- Slide 12
- Slide 13
- Slide 14
- Slide 15
- Slide 16
- Slide 17
- Slide 18
- Slide 19
- Slide 20
- Slide 21
- Slide 22
- Slide 23
- Slide 24
- Slide 25
- Slide 26
- Slide 27
- Slide 28
- Slide 29
- Slide 30
- Slide 31
- Slide 32
- Slide 33
- Slide 34
- Slide 35
- Slide 36
- Slide 37
- Slide 38
- Slide 39
- Slide 40
- Slide 41
- Slide 42
- Slide 43
- Slide 44
- Slide 45
- Slide 46
- Slide 47
- Slide 48
- Announcements
- Slide 50
-
bull Ask the user explicitly learnbull NLP intelligent dialoguesbull Bayesian networks Hidden Markov models
bull Observe the user learn bull Logs machine learningbull Clustering classification data mining
bull Interactive user modeling mixture of direct inputs of a user observations and inferences
User Model Elicitation
Social Web 2016 Lora Aroyo and Davide Ceolin
httphunchcomSocial Web 2016 Lora Aroyo and Davide Ceolin
Social Web 2016 Lora Aroyo and Davide Ceolin
httpfarm1staticflickrcom155413650229_31ef379b0b_bjpg
bull set of characteristics (eg attribute-value pairs) that describe a group of users
bull user is not assigned to a single stereotype - user profile can feature characteristics of several different stereotypes
User Stereotypes
based on slides from Fabien Abel
based on slides from Fabien Abel
Can we infer a Twitter-based User
Profile
User Modeling Building Blocks
based on slides from Fabien Abel
User Modeling Building Blocks
based on slides from Fabien Abel
User Modeling Building Blocks
based on slides from Fabien Abel
User Modeling Building Blocks
based on slides from Fabien Abel
Observationsbull Profile characteristics
bull Semantic enrichment solves sparsity problemsbull Profiles change over time recent profiles reflect
better current user demandsbull Temporal patterns weekend profiles differ
significantly from weekday profilesbull Impact on recommendations
bull The more fine-grained the concepts the better the recommendation performance entity-based gt topic-based gt hashtag-based
bull Semantic enrichment improves recommendation quality
bull Time-sensitivity (adapting to trends) improves performance
Social Web 2016 Lora Aroyo and Davide Ceolin
User Modelingit is not about putting everything in a user profile
it is about making the right choices
Social Web 2016 Lora Aroyo and Davide Ceolin
User Adaptation
Knowing the user to adapt a system or interfaceto improve the system functionality and user
experience
Social Web 2016 Lora Aroyo and Davide Ceolin
A Jameson Adaptive interfaces and agents The HCI handbook fundamentals evolving technologies and emerging applications pp 305ndash330 2003
User-Adaptive Systems
Social Web 2016 Lora Aroyo and Davide Ceolin
based on slides from Fabien Abel
Lastfm adapts to your music taste
bull Overfitting ldquobubble effectsrdquo loss of serendipity problem bull systems may adapt too strongly to the
interestsbehaviorbull eg an adaptive radio station may always play the
same or very similar songsbull We search for the right balance between novelty and
relevance for the userbull ldquoLost in Hyperspacerdquo problem
bull when adapting the navigation ndash ie the links on which users can click to findaccess information
bull eg re-orderinghiding of menu items may lead to confusion
Issues in User-Adaptive Systems
Social Web 2016 Lora Aroyo and Davide Ceolin
What is good user modelling amp
personalisation
httpwwwflickrcomphotosbellarosebyliz4729613108
From the consumer perspective of an adaptive system
From the provider perspective of an adaptive system
Success Perspectives
Social Web 2016 Lora Aroyo and Davide Ceolin
bull User studies askobserve (selected) people whether you did a good job
bull Log analysis Analyze (click) data and infer whether you did a good job
bull Evaluation of user modelingbull measure quality of profiles directly eg measure
overlap with existing (true) profiles or let people judge the quality of the generated user profiles
bull measure quality of application that exploits the user profile eg apply user modeling strategies in a recommender system
Evaluation Strategies
Social Web 2016 Lora Aroyo and Davide Ceolin
Evaluating User Modeling in RecSys
Social Web 2016 Lora Aroyo and Davide Ceolin
Possible Metrics
Social Web 2016 Lora Aroyo and Davide Ceolin
bull The usual IR metricsbull Precision fraction of retrieved items that are relevantbull Recall fraction of relevant items that have been
retrievedbull F-Measure (harmonic) mean of precision and recall
bull Metrics for evaluating recommendation (rankings)bull Mean Reciprocal Rank (MRR) of first relevant itembull Successk probability that relevant item occurs
within the top kbull If a true ranking is given rank correlations bull Precisionk Recallk amp F-Measurek
bull Metrics for evaluating prediction of user preferencesbull MAE = Mean Absolute Errorbull TrueFalse PositivesNegatives
Predict relevantusefulinteresting items for a given user (in a given context) itrsquos often a ranking task
Recommendation Systems
Social Web 2016 Lora Aroyo and Davide Ceolin
March 28 2013
Social Web 2016 Lora Aroyo and Davide Ceolin
Social Web 2016 Lora Aroyo and Davide Ceolin
Social Web 2016 Lora Aroyo and Davide Ceolin
httpwwwwiredcommagazine201111mf_artsyall1
Social Web 2016 Lora Aroyo and Davide Ceolin
Collaborative Filteringbull Memory-based User-Item matrix ratingspreferences of users =gt
compute similarity between users amp recommend items of similar users
bull Model-based Item-Item matrix similarity (eg based on user ratings) between items =gt recommend items that are similar to the ones the user likes
bull Model-based Clustering cluster users according to their preferences =gt recommend items of users that belong to the same cluster
bull Model-based Bayesian networks P(u likes item B | u likes item A) = how likely is it that a user who likes item A will like item B learn probabilities from user ratingspreferences
bull Others rule-based other data mining techniques
Social Web 2016 Lora Aroyo and Davide Ceolin
bull complete input data is required
bull pre-computation not possible
bull does not scale well bull high quality of
recommendations
bull abstraction (model) of input data
bull pre-computation (partially) possible (model has to be re-built from time to time)
bull scales betterbull abstraction may reduce
recommendation quality
Memory vs Model-based
Social Web 2016 Lora Aroyo and Davide Ceolin
bull collaborative filtering lsquoneighborhoodsrsquo of people with similar interest amp recommending items based on likings in neighborhood
bull limitations next to lsquocold startrsquo and lsquosparsityrsquo the lack of control (over onersquos neighborhood) is also a problem ie cannot add lsquotrustedrsquo people nor exclude lsquostrangersquo ones
bull therefore interest in lsquosocial recommendersrsquo where presence of social connections defines the similarity in interests (eg social tagging CiteULike)bull does a social connection indicate user interest similaritybull how much users interest similarity depends on the strength
of their connectionbull is it feasible to use a social network as a personalized
recommendation[Lin amp Brusilovsky Social Networks and Interest Similarity The Case of CiteULike HTrsquo10]
Social Networks amp Interest Similarity
bull unilaterally connected pairs have more common itemsmetadatatags than non-connected pairsbull highest similarity for direct connections - decreasing with the increase of distance between users in SN bull reciprocal relationship users - significantly larger similarity than users in a unidirectional relationship bull traditional item-level similarity may be less reliable to find similar users in social bookmarking systemsbull peers connected by self-defined social connections could be a useful source for cross-recommendation
Conclusions
Social Web 2016 Lora Aroyo and Davide Ceolin
bull Input characteristics of items amp interests of a user into characteristics of items =gt Recommend items that feature characteristics which meet the userrsquos interests
bull Techniquesbull Data mining methods Cluster items based on their
characteristics =gt Infer usersrsquo interests into clustersbull IR methods Represent items amp users as term vectors
=gt Compute similarity between user profile vector and items
bull Utility-based methods Utility function that gets an item as input the parameters of the utility function are customized via preferences of a user
Content-based Recommendations
Social Web 2016 Lora Aroyo and Davide Ceolin
based on slides from Fabien Abel
Content Features
based on slides from Fabien Abel
User Model
based on slides from Fabien Abel
Recommendations
RecSys Issuesbull Cold-start problem (new user problem) nolittle data available to infer preferences of new
usersbull Changing User Preferences user interests may change over timebull Sparsity problem (new item problem) item descriptions are sparse eg not many user
rated or tagged an itembull Lack of Diversity (overfitting) when adapting too strongly to the preferences of users they
might see samesimilar recommendationsbull Use the right context users do things which might not be relevant for their user model
eg try out things do stuff for other peoplebull Research challenge right balance between serendipity amp personalizationbull Research challenge right way to use the influence of recommendations on userrsquos behavior
Social Web 2016 Lora Aroyo and Davide Ceolin
Announcementsbull Next deadlinesbull Tuesday March 1st 2359
Assignment 2bull Friday March 4th 1000 Post
questionbull Friday March 4th 1700 Vote
questions
image source httpwwwflickrcomphotosbionicteaching1375254387
Hands-on Teaser
bull Your Facebook Friendsrsquo popularity in a spread sheet
bull Locations of your Facebook Friendsbull Tag Cloud of your wall posts
- Social Web 2016
- Slide 2
- Slide 3
- Slide 4
- Slide 5
- Slide 6
- Slide 7
- Slide 8
- Slide 9
- Slide 10
- Slide 11
- Slide 12
- Slide 13
- Slide 14
- Slide 15
- Slide 16
- Slide 17
- Slide 18
- Slide 19
- Slide 20
- Slide 21
- Slide 22
- Slide 23
- Slide 24
- Slide 25
- Slide 26
- Slide 27
- Slide 28
- Slide 29
- Slide 30
- Slide 31
- Slide 32
- Slide 33
- Slide 34
- Slide 35
- Slide 36
- Slide 37
- Slide 38
- Slide 39
- Slide 40
- Slide 41
- Slide 42
- Slide 43
- Slide 44
- Slide 45
- Slide 46
- Slide 47
- Slide 48
- Announcements
- Slide 50
-
httphunchcomSocial Web 2016 Lora Aroyo and Davide Ceolin
Social Web 2016 Lora Aroyo and Davide Ceolin
httpfarm1staticflickrcom155413650229_31ef379b0b_bjpg
bull set of characteristics (eg attribute-value pairs) that describe a group of users
bull user is not assigned to a single stereotype - user profile can feature characteristics of several different stereotypes
User Stereotypes
based on slides from Fabien Abel
based on slides from Fabien Abel
Can we infer a Twitter-based User
Profile
User Modeling Building Blocks
based on slides from Fabien Abel
User Modeling Building Blocks
based on slides from Fabien Abel
User Modeling Building Blocks
based on slides from Fabien Abel
User Modeling Building Blocks
based on slides from Fabien Abel
Observationsbull Profile characteristics
bull Semantic enrichment solves sparsity problemsbull Profiles change over time recent profiles reflect
better current user demandsbull Temporal patterns weekend profiles differ
significantly from weekday profilesbull Impact on recommendations
bull The more fine-grained the concepts the better the recommendation performance entity-based gt topic-based gt hashtag-based
bull Semantic enrichment improves recommendation quality
bull Time-sensitivity (adapting to trends) improves performance
Social Web 2016 Lora Aroyo and Davide Ceolin
User Modelingit is not about putting everything in a user profile
it is about making the right choices
Social Web 2016 Lora Aroyo and Davide Ceolin
User Adaptation
Knowing the user to adapt a system or interfaceto improve the system functionality and user
experience
Social Web 2016 Lora Aroyo and Davide Ceolin
A Jameson Adaptive interfaces and agents The HCI handbook fundamentals evolving technologies and emerging applications pp 305ndash330 2003
User-Adaptive Systems
Social Web 2016 Lora Aroyo and Davide Ceolin
based on slides from Fabien Abel
Lastfm adapts to your music taste
bull Overfitting ldquobubble effectsrdquo loss of serendipity problem bull systems may adapt too strongly to the
interestsbehaviorbull eg an adaptive radio station may always play the
same or very similar songsbull We search for the right balance between novelty and
relevance for the userbull ldquoLost in Hyperspacerdquo problem
bull when adapting the navigation ndash ie the links on which users can click to findaccess information
bull eg re-orderinghiding of menu items may lead to confusion
Issues in User-Adaptive Systems
Social Web 2016 Lora Aroyo and Davide Ceolin
What is good user modelling amp
personalisation
httpwwwflickrcomphotosbellarosebyliz4729613108
From the consumer perspective of an adaptive system
From the provider perspective of an adaptive system
Success Perspectives
Social Web 2016 Lora Aroyo and Davide Ceolin
bull User studies askobserve (selected) people whether you did a good job
bull Log analysis Analyze (click) data and infer whether you did a good job
bull Evaluation of user modelingbull measure quality of profiles directly eg measure
overlap with existing (true) profiles or let people judge the quality of the generated user profiles
bull measure quality of application that exploits the user profile eg apply user modeling strategies in a recommender system
Evaluation Strategies
Social Web 2016 Lora Aroyo and Davide Ceolin
Evaluating User Modeling in RecSys
Social Web 2016 Lora Aroyo and Davide Ceolin
Possible Metrics
Social Web 2016 Lora Aroyo and Davide Ceolin
bull The usual IR metricsbull Precision fraction of retrieved items that are relevantbull Recall fraction of relevant items that have been
retrievedbull F-Measure (harmonic) mean of precision and recall
bull Metrics for evaluating recommendation (rankings)bull Mean Reciprocal Rank (MRR) of first relevant itembull Successk probability that relevant item occurs
within the top kbull If a true ranking is given rank correlations bull Precisionk Recallk amp F-Measurek
bull Metrics for evaluating prediction of user preferencesbull MAE = Mean Absolute Errorbull TrueFalse PositivesNegatives
Predict relevantusefulinteresting items for a given user (in a given context) itrsquos often a ranking task
Recommendation Systems
Social Web 2016 Lora Aroyo and Davide Ceolin
March 28 2013
Social Web 2016 Lora Aroyo and Davide Ceolin
Social Web 2016 Lora Aroyo and Davide Ceolin
Social Web 2016 Lora Aroyo and Davide Ceolin
httpwwwwiredcommagazine201111mf_artsyall1
Social Web 2016 Lora Aroyo and Davide Ceolin
Collaborative Filteringbull Memory-based User-Item matrix ratingspreferences of users =gt
compute similarity between users amp recommend items of similar users
bull Model-based Item-Item matrix similarity (eg based on user ratings) between items =gt recommend items that are similar to the ones the user likes
bull Model-based Clustering cluster users according to their preferences =gt recommend items of users that belong to the same cluster
bull Model-based Bayesian networks P(u likes item B | u likes item A) = how likely is it that a user who likes item A will like item B learn probabilities from user ratingspreferences
bull Others rule-based other data mining techniques
Social Web 2016 Lora Aroyo and Davide Ceolin
bull complete input data is required
bull pre-computation not possible
bull does not scale well bull high quality of
recommendations
bull abstraction (model) of input data
bull pre-computation (partially) possible (model has to be re-built from time to time)
bull scales betterbull abstraction may reduce
recommendation quality
Memory vs Model-based
Social Web 2016 Lora Aroyo and Davide Ceolin
bull collaborative filtering lsquoneighborhoodsrsquo of people with similar interest amp recommending items based on likings in neighborhood
bull limitations next to lsquocold startrsquo and lsquosparsityrsquo the lack of control (over onersquos neighborhood) is also a problem ie cannot add lsquotrustedrsquo people nor exclude lsquostrangersquo ones
bull therefore interest in lsquosocial recommendersrsquo where presence of social connections defines the similarity in interests (eg social tagging CiteULike)bull does a social connection indicate user interest similaritybull how much users interest similarity depends on the strength
of their connectionbull is it feasible to use a social network as a personalized
recommendation[Lin amp Brusilovsky Social Networks and Interest Similarity The Case of CiteULike HTrsquo10]
Social Networks amp Interest Similarity
bull unilaterally connected pairs have more common itemsmetadatatags than non-connected pairsbull highest similarity for direct connections - decreasing with the increase of distance between users in SN bull reciprocal relationship users - significantly larger similarity than users in a unidirectional relationship bull traditional item-level similarity may be less reliable to find similar users in social bookmarking systemsbull peers connected by self-defined social connections could be a useful source for cross-recommendation
Conclusions
Social Web 2016 Lora Aroyo and Davide Ceolin
bull Input characteristics of items amp interests of a user into characteristics of items =gt Recommend items that feature characteristics which meet the userrsquos interests
bull Techniquesbull Data mining methods Cluster items based on their
characteristics =gt Infer usersrsquo interests into clustersbull IR methods Represent items amp users as term vectors
=gt Compute similarity between user profile vector and items
bull Utility-based methods Utility function that gets an item as input the parameters of the utility function are customized via preferences of a user
Content-based Recommendations
Social Web 2016 Lora Aroyo and Davide Ceolin
based on slides from Fabien Abel
Content Features
based on slides from Fabien Abel
User Model
based on slides from Fabien Abel
Recommendations
RecSys Issuesbull Cold-start problem (new user problem) nolittle data available to infer preferences of new
usersbull Changing User Preferences user interests may change over timebull Sparsity problem (new item problem) item descriptions are sparse eg not many user
rated or tagged an itembull Lack of Diversity (overfitting) when adapting too strongly to the preferences of users they
might see samesimilar recommendationsbull Use the right context users do things which might not be relevant for their user model
eg try out things do stuff for other peoplebull Research challenge right balance between serendipity amp personalizationbull Research challenge right way to use the influence of recommendations on userrsquos behavior
Social Web 2016 Lora Aroyo and Davide Ceolin
Announcementsbull Next deadlinesbull Tuesday March 1st 2359
Assignment 2bull Friday March 4th 1000 Post
questionbull Friday March 4th 1700 Vote
questions
image source httpwwwflickrcomphotosbionicteaching1375254387
Hands-on Teaser
bull Your Facebook Friendsrsquo popularity in a spread sheet
bull Locations of your Facebook Friendsbull Tag Cloud of your wall posts
- Social Web 2016
- Slide 2
- Slide 3
- Slide 4
- Slide 5
- Slide 6
- Slide 7
- Slide 8
- Slide 9
- Slide 10
- Slide 11
- Slide 12
- Slide 13
- Slide 14
- Slide 15
- Slide 16
- Slide 17
- Slide 18
- Slide 19
- Slide 20
- Slide 21
- Slide 22
- Slide 23
- Slide 24
- Slide 25
- Slide 26
- Slide 27
- Slide 28
- Slide 29
- Slide 30
- Slide 31
- Slide 32
- Slide 33
- Slide 34
- Slide 35
- Slide 36
- Slide 37
- Slide 38
- Slide 39
- Slide 40
- Slide 41
- Slide 42
- Slide 43
- Slide 44
- Slide 45
- Slide 46
- Slide 47
- Slide 48
- Announcements
- Slide 50
-
Social Web 2016 Lora Aroyo and Davide Ceolin
httpfarm1staticflickrcom155413650229_31ef379b0b_bjpg
bull set of characteristics (eg attribute-value pairs) that describe a group of users
bull user is not assigned to a single stereotype - user profile can feature characteristics of several different stereotypes
User Stereotypes
based on slides from Fabien Abel
based on slides from Fabien Abel
Can we infer a Twitter-based User
Profile
User Modeling Building Blocks
based on slides from Fabien Abel
User Modeling Building Blocks
based on slides from Fabien Abel
User Modeling Building Blocks
based on slides from Fabien Abel
User Modeling Building Blocks
based on slides from Fabien Abel
Observationsbull Profile characteristics
bull Semantic enrichment solves sparsity problemsbull Profiles change over time recent profiles reflect
better current user demandsbull Temporal patterns weekend profiles differ
significantly from weekday profilesbull Impact on recommendations
bull The more fine-grained the concepts the better the recommendation performance entity-based gt topic-based gt hashtag-based
bull Semantic enrichment improves recommendation quality
bull Time-sensitivity (adapting to trends) improves performance
Social Web 2016 Lora Aroyo and Davide Ceolin
User Modelingit is not about putting everything in a user profile
it is about making the right choices
Social Web 2016 Lora Aroyo and Davide Ceolin
User Adaptation
Knowing the user to adapt a system or interfaceto improve the system functionality and user
experience
Social Web 2016 Lora Aroyo and Davide Ceolin
A Jameson Adaptive interfaces and agents The HCI handbook fundamentals evolving technologies and emerging applications pp 305ndash330 2003
User-Adaptive Systems
Social Web 2016 Lora Aroyo and Davide Ceolin
based on slides from Fabien Abel
Lastfm adapts to your music taste
bull Overfitting ldquobubble effectsrdquo loss of serendipity problem bull systems may adapt too strongly to the
interestsbehaviorbull eg an adaptive radio station may always play the
same or very similar songsbull We search for the right balance between novelty and
relevance for the userbull ldquoLost in Hyperspacerdquo problem
bull when adapting the navigation ndash ie the links on which users can click to findaccess information
bull eg re-orderinghiding of menu items may lead to confusion
Issues in User-Adaptive Systems
Social Web 2016 Lora Aroyo and Davide Ceolin
What is good user modelling amp
personalisation
httpwwwflickrcomphotosbellarosebyliz4729613108
From the consumer perspective of an adaptive system
From the provider perspective of an adaptive system
Success Perspectives
Social Web 2016 Lora Aroyo and Davide Ceolin
bull User studies askobserve (selected) people whether you did a good job
bull Log analysis Analyze (click) data and infer whether you did a good job
bull Evaluation of user modelingbull measure quality of profiles directly eg measure
overlap with existing (true) profiles or let people judge the quality of the generated user profiles
bull measure quality of application that exploits the user profile eg apply user modeling strategies in a recommender system
Evaluation Strategies
Social Web 2016 Lora Aroyo and Davide Ceolin
Evaluating User Modeling in RecSys
Social Web 2016 Lora Aroyo and Davide Ceolin
Possible Metrics
Social Web 2016 Lora Aroyo and Davide Ceolin
bull The usual IR metricsbull Precision fraction of retrieved items that are relevantbull Recall fraction of relevant items that have been
retrievedbull F-Measure (harmonic) mean of precision and recall
bull Metrics for evaluating recommendation (rankings)bull Mean Reciprocal Rank (MRR) of first relevant itembull Successk probability that relevant item occurs
within the top kbull If a true ranking is given rank correlations bull Precisionk Recallk amp F-Measurek
bull Metrics for evaluating prediction of user preferencesbull MAE = Mean Absolute Errorbull TrueFalse PositivesNegatives
Predict relevantusefulinteresting items for a given user (in a given context) itrsquos often a ranking task
Recommendation Systems
Social Web 2016 Lora Aroyo and Davide Ceolin
March 28 2013
Social Web 2016 Lora Aroyo and Davide Ceolin
Social Web 2016 Lora Aroyo and Davide Ceolin
Social Web 2016 Lora Aroyo and Davide Ceolin
httpwwwwiredcommagazine201111mf_artsyall1
Social Web 2016 Lora Aroyo and Davide Ceolin
Collaborative Filteringbull Memory-based User-Item matrix ratingspreferences of users =gt
compute similarity between users amp recommend items of similar users
bull Model-based Item-Item matrix similarity (eg based on user ratings) between items =gt recommend items that are similar to the ones the user likes
bull Model-based Clustering cluster users according to their preferences =gt recommend items of users that belong to the same cluster
bull Model-based Bayesian networks P(u likes item B | u likes item A) = how likely is it that a user who likes item A will like item B learn probabilities from user ratingspreferences
bull Others rule-based other data mining techniques
Social Web 2016 Lora Aroyo and Davide Ceolin
bull complete input data is required
bull pre-computation not possible
bull does not scale well bull high quality of
recommendations
bull abstraction (model) of input data
bull pre-computation (partially) possible (model has to be re-built from time to time)
bull scales betterbull abstraction may reduce
recommendation quality
Memory vs Model-based
Social Web 2016 Lora Aroyo and Davide Ceolin
bull collaborative filtering lsquoneighborhoodsrsquo of people with similar interest amp recommending items based on likings in neighborhood
bull limitations next to lsquocold startrsquo and lsquosparsityrsquo the lack of control (over onersquos neighborhood) is also a problem ie cannot add lsquotrustedrsquo people nor exclude lsquostrangersquo ones
bull therefore interest in lsquosocial recommendersrsquo where presence of social connections defines the similarity in interests (eg social tagging CiteULike)bull does a social connection indicate user interest similaritybull how much users interest similarity depends on the strength
of their connectionbull is it feasible to use a social network as a personalized
recommendation[Lin amp Brusilovsky Social Networks and Interest Similarity The Case of CiteULike HTrsquo10]
Social Networks amp Interest Similarity
bull unilaterally connected pairs have more common itemsmetadatatags than non-connected pairsbull highest similarity for direct connections - decreasing with the increase of distance between users in SN bull reciprocal relationship users - significantly larger similarity than users in a unidirectional relationship bull traditional item-level similarity may be less reliable to find similar users in social bookmarking systemsbull peers connected by self-defined social connections could be a useful source for cross-recommendation
Conclusions
Social Web 2016 Lora Aroyo and Davide Ceolin
bull Input characteristics of items amp interests of a user into characteristics of items =gt Recommend items that feature characteristics which meet the userrsquos interests
bull Techniquesbull Data mining methods Cluster items based on their
characteristics =gt Infer usersrsquo interests into clustersbull IR methods Represent items amp users as term vectors
=gt Compute similarity between user profile vector and items
bull Utility-based methods Utility function that gets an item as input the parameters of the utility function are customized via preferences of a user
Content-based Recommendations
Social Web 2016 Lora Aroyo and Davide Ceolin
based on slides from Fabien Abel
Content Features
based on slides from Fabien Abel
User Model
based on slides from Fabien Abel
Recommendations
RecSys Issuesbull Cold-start problem (new user problem) nolittle data available to infer preferences of new
usersbull Changing User Preferences user interests may change over timebull Sparsity problem (new item problem) item descriptions are sparse eg not many user
rated or tagged an itembull Lack of Diversity (overfitting) when adapting too strongly to the preferences of users they
might see samesimilar recommendationsbull Use the right context users do things which might not be relevant for their user model
eg try out things do stuff for other peoplebull Research challenge right balance between serendipity amp personalizationbull Research challenge right way to use the influence of recommendations on userrsquos behavior
Social Web 2016 Lora Aroyo and Davide Ceolin
Announcementsbull Next deadlinesbull Tuesday March 1st 2359
Assignment 2bull Friday March 4th 1000 Post
questionbull Friday March 4th 1700 Vote
questions
image source httpwwwflickrcomphotosbionicteaching1375254387
Hands-on Teaser
bull Your Facebook Friendsrsquo popularity in a spread sheet
bull Locations of your Facebook Friendsbull Tag Cloud of your wall posts
- Social Web 2016
- Slide 2
- Slide 3
- Slide 4
- Slide 5
- Slide 6
- Slide 7
- Slide 8
- Slide 9
- Slide 10
- Slide 11
- Slide 12
- Slide 13
- Slide 14
- Slide 15
- Slide 16
- Slide 17
- Slide 18
- Slide 19
- Slide 20
- Slide 21
- Slide 22
- Slide 23
- Slide 24
- Slide 25
- Slide 26
- Slide 27
- Slide 28
- Slide 29
- Slide 30
- Slide 31
- Slide 32
- Slide 33
- Slide 34
- Slide 35
- Slide 36
- Slide 37
- Slide 38
- Slide 39
- Slide 40
- Slide 41
- Slide 42
- Slide 43
- Slide 44
- Slide 45
- Slide 46
- Slide 47
- Slide 48
- Announcements
- Slide 50
-
httpfarm1staticflickrcom155413650229_31ef379b0b_bjpg
bull set of characteristics (eg attribute-value pairs) that describe a group of users
bull user is not assigned to a single stereotype - user profile can feature characteristics of several different stereotypes
User Stereotypes
based on slides from Fabien Abel
based on slides from Fabien Abel
Can we infer a Twitter-based User
Profile
User Modeling Building Blocks
based on slides from Fabien Abel
User Modeling Building Blocks
based on slides from Fabien Abel
User Modeling Building Blocks
based on slides from Fabien Abel
User Modeling Building Blocks
based on slides from Fabien Abel
Observationsbull Profile characteristics
bull Semantic enrichment solves sparsity problemsbull Profiles change over time recent profiles reflect
better current user demandsbull Temporal patterns weekend profiles differ
significantly from weekday profilesbull Impact on recommendations
bull The more fine-grained the concepts the better the recommendation performance entity-based gt topic-based gt hashtag-based
bull Semantic enrichment improves recommendation quality
bull Time-sensitivity (adapting to trends) improves performance
Social Web 2016 Lora Aroyo and Davide Ceolin
User Modelingit is not about putting everything in a user profile
it is about making the right choices
Social Web 2016 Lora Aroyo and Davide Ceolin
User Adaptation
Knowing the user to adapt a system or interfaceto improve the system functionality and user
experience
Social Web 2016 Lora Aroyo and Davide Ceolin
A Jameson Adaptive interfaces and agents The HCI handbook fundamentals evolving technologies and emerging applications pp 305ndash330 2003
User-Adaptive Systems
Social Web 2016 Lora Aroyo and Davide Ceolin
based on slides from Fabien Abel
Lastfm adapts to your music taste
bull Overfitting ldquobubble effectsrdquo loss of serendipity problem bull systems may adapt too strongly to the
interestsbehaviorbull eg an adaptive radio station may always play the
same or very similar songsbull We search for the right balance between novelty and
relevance for the userbull ldquoLost in Hyperspacerdquo problem
bull when adapting the navigation ndash ie the links on which users can click to findaccess information
bull eg re-orderinghiding of menu items may lead to confusion
Issues in User-Adaptive Systems
Social Web 2016 Lora Aroyo and Davide Ceolin
What is good user modelling amp
personalisation
httpwwwflickrcomphotosbellarosebyliz4729613108
From the consumer perspective of an adaptive system
From the provider perspective of an adaptive system
Success Perspectives
Social Web 2016 Lora Aroyo and Davide Ceolin
bull User studies askobserve (selected) people whether you did a good job
bull Log analysis Analyze (click) data and infer whether you did a good job
bull Evaluation of user modelingbull measure quality of profiles directly eg measure
overlap with existing (true) profiles or let people judge the quality of the generated user profiles
bull measure quality of application that exploits the user profile eg apply user modeling strategies in a recommender system
Evaluation Strategies
Social Web 2016 Lora Aroyo and Davide Ceolin
Evaluating User Modeling in RecSys
Social Web 2016 Lora Aroyo and Davide Ceolin
Possible Metrics
Social Web 2016 Lora Aroyo and Davide Ceolin
bull The usual IR metricsbull Precision fraction of retrieved items that are relevantbull Recall fraction of relevant items that have been
retrievedbull F-Measure (harmonic) mean of precision and recall
bull Metrics for evaluating recommendation (rankings)bull Mean Reciprocal Rank (MRR) of first relevant itembull Successk probability that relevant item occurs
within the top kbull If a true ranking is given rank correlations bull Precisionk Recallk amp F-Measurek
bull Metrics for evaluating prediction of user preferencesbull MAE = Mean Absolute Errorbull TrueFalse PositivesNegatives
Predict relevantusefulinteresting items for a given user (in a given context) itrsquos often a ranking task
Recommendation Systems
Social Web 2016 Lora Aroyo and Davide Ceolin
March 28 2013
Social Web 2016 Lora Aroyo and Davide Ceolin
Social Web 2016 Lora Aroyo and Davide Ceolin
Social Web 2016 Lora Aroyo and Davide Ceolin
httpwwwwiredcommagazine201111mf_artsyall1
Social Web 2016 Lora Aroyo and Davide Ceolin
Collaborative Filteringbull Memory-based User-Item matrix ratingspreferences of users =gt
compute similarity between users amp recommend items of similar users
bull Model-based Item-Item matrix similarity (eg based on user ratings) between items =gt recommend items that are similar to the ones the user likes
bull Model-based Clustering cluster users according to their preferences =gt recommend items of users that belong to the same cluster
bull Model-based Bayesian networks P(u likes item B | u likes item A) = how likely is it that a user who likes item A will like item B learn probabilities from user ratingspreferences
bull Others rule-based other data mining techniques
Social Web 2016 Lora Aroyo and Davide Ceolin
bull complete input data is required
bull pre-computation not possible
bull does not scale well bull high quality of
recommendations
bull abstraction (model) of input data
bull pre-computation (partially) possible (model has to be re-built from time to time)
bull scales betterbull abstraction may reduce
recommendation quality
Memory vs Model-based
Social Web 2016 Lora Aroyo and Davide Ceolin
bull collaborative filtering lsquoneighborhoodsrsquo of people with similar interest amp recommending items based on likings in neighborhood
bull limitations next to lsquocold startrsquo and lsquosparsityrsquo the lack of control (over onersquos neighborhood) is also a problem ie cannot add lsquotrustedrsquo people nor exclude lsquostrangersquo ones
bull therefore interest in lsquosocial recommendersrsquo where presence of social connections defines the similarity in interests (eg social tagging CiteULike)bull does a social connection indicate user interest similaritybull how much users interest similarity depends on the strength
of their connectionbull is it feasible to use a social network as a personalized
recommendation[Lin amp Brusilovsky Social Networks and Interest Similarity The Case of CiteULike HTrsquo10]
Social Networks amp Interest Similarity
bull unilaterally connected pairs have more common itemsmetadatatags than non-connected pairsbull highest similarity for direct connections - decreasing with the increase of distance between users in SN bull reciprocal relationship users - significantly larger similarity than users in a unidirectional relationship bull traditional item-level similarity may be less reliable to find similar users in social bookmarking systemsbull peers connected by self-defined social connections could be a useful source for cross-recommendation
Conclusions
Social Web 2016 Lora Aroyo and Davide Ceolin
bull Input characteristics of items amp interests of a user into characteristics of items =gt Recommend items that feature characteristics which meet the userrsquos interests
bull Techniquesbull Data mining methods Cluster items based on their
characteristics =gt Infer usersrsquo interests into clustersbull IR methods Represent items amp users as term vectors
=gt Compute similarity between user profile vector and items
bull Utility-based methods Utility function that gets an item as input the parameters of the utility function are customized via preferences of a user
Content-based Recommendations
Social Web 2016 Lora Aroyo and Davide Ceolin
based on slides from Fabien Abel
Content Features
based on slides from Fabien Abel
User Model
based on slides from Fabien Abel
Recommendations
RecSys Issuesbull Cold-start problem (new user problem) nolittle data available to infer preferences of new
usersbull Changing User Preferences user interests may change over timebull Sparsity problem (new item problem) item descriptions are sparse eg not many user
rated or tagged an itembull Lack of Diversity (overfitting) when adapting too strongly to the preferences of users they
might see samesimilar recommendationsbull Use the right context users do things which might not be relevant for their user model
eg try out things do stuff for other peoplebull Research challenge right balance between serendipity amp personalizationbull Research challenge right way to use the influence of recommendations on userrsquos behavior
Social Web 2016 Lora Aroyo and Davide Ceolin
Announcementsbull Next deadlinesbull Tuesday March 1st 2359
Assignment 2bull Friday March 4th 1000 Post
questionbull Friday March 4th 1700 Vote
questions
image source httpwwwflickrcomphotosbionicteaching1375254387
Hands-on Teaser
bull Your Facebook Friendsrsquo popularity in a spread sheet
bull Locations of your Facebook Friendsbull Tag Cloud of your wall posts
- Social Web 2016
- Slide 2
- Slide 3
- Slide 4
- Slide 5
- Slide 6
- Slide 7
- Slide 8
- Slide 9
- Slide 10
- Slide 11
- Slide 12
- Slide 13
- Slide 14
- Slide 15
- Slide 16
- Slide 17
- Slide 18
- Slide 19
- Slide 20
- Slide 21
- Slide 22
- Slide 23
- Slide 24
- Slide 25
- Slide 26
- Slide 27
- Slide 28
- Slide 29
- Slide 30
- Slide 31
- Slide 32
- Slide 33
- Slide 34
- Slide 35
- Slide 36
- Slide 37
- Slide 38
- Slide 39
- Slide 40
- Slide 41
- Slide 42
- Slide 43
- Slide 44
- Slide 45
- Slide 46
- Slide 47
- Slide 48
- Announcements
- Slide 50
-
based on slides from Fabien Abel
based on slides from Fabien Abel
Can we infer a Twitter-based User
Profile
User Modeling Building Blocks
based on slides from Fabien Abel
User Modeling Building Blocks
based on slides from Fabien Abel
User Modeling Building Blocks
based on slides from Fabien Abel
User Modeling Building Blocks
based on slides from Fabien Abel
Observationsbull Profile characteristics
bull Semantic enrichment solves sparsity problemsbull Profiles change over time recent profiles reflect
better current user demandsbull Temporal patterns weekend profiles differ
significantly from weekday profilesbull Impact on recommendations
bull The more fine-grained the concepts the better the recommendation performance entity-based gt topic-based gt hashtag-based
bull Semantic enrichment improves recommendation quality
bull Time-sensitivity (adapting to trends) improves performance
Social Web 2016 Lora Aroyo and Davide Ceolin
User Modelingit is not about putting everything in a user profile
it is about making the right choices
Social Web 2016 Lora Aroyo and Davide Ceolin
User Adaptation
Knowing the user to adapt a system or interfaceto improve the system functionality and user
experience
Social Web 2016 Lora Aroyo and Davide Ceolin
A Jameson Adaptive interfaces and agents The HCI handbook fundamentals evolving technologies and emerging applications pp 305ndash330 2003
User-Adaptive Systems
Social Web 2016 Lora Aroyo and Davide Ceolin
based on slides from Fabien Abel
Lastfm adapts to your music taste
bull Overfitting ldquobubble effectsrdquo loss of serendipity problem bull systems may adapt too strongly to the
interestsbehaviorbull eg an adaptive radio station may always play the
same or very similar songsbull We search for the right balance between novelty and
relevance for the userbull ldquoLost in Hyperspacerdquo problem
bull when adapting the navigation ndash ie the links on which users can click to findaccess information
bull eg re-orderinghiding of menu items may lead to confusion
Issues in User-Adaptive Systems
Social Web 2016 Lora Aroyo and Davide Ceolin
What is good user modelling amp
personalisation
httpwwwflickrcomphotosbellarosebyliz4729613108
From the consumer perspective of an adaptive system
From the provider perspective of an adaptive system
Success Perspectives
Social Web 2016 Lora Aroyo and Davide Ceolin
bull User studies askobserve (selected) people whether you did a good job
bull Log analysis Analyze (click) data and infer whether you did a good job
bull Evaluation of user modelingbull measure quality of profiles directly eg measure
overlap with existing (true) profiles or let people judge the quality of the generated user profiles
bull measure quality of application that exploits the user profile eg apply user modeling strategies in a recommender system
Evaluation Strategies
Social Web 2016 Lora Aroyo and Davide Ceolin
Evaluating User Modeling in RecSys
Social Web 2016 Lora Aroyo and Davide Ceolin
Possible Metrics
Social Web 2016 Lora Aroyo and Davide Ceolin
bull The usual IR metricsbull Precision fraction of retrieved items that are relevantbull Recall fraction of relevant items that have been
retrievedbull F-Measure (harmonic) mean of precision and recall
bull Metrics for evaluating recommendation (rankings)bull Mean Reciprocal Rank (MRR) of first relevant itembull Successk probability that relevant item occurs
within the top kbull If a true ranking is given rank correlations bull Precisionk Recallk amp F-Measurek
bull Metrics for evaluating prediction of user preferencesbull MAE = Mean Absolute Errorbull TrueFalse PositivesNegatives
Predict relevantusefulinteresting items for a given user (in a given context) itrsquos often a ranking task
Recommendation Systems
Social Web 2016 Lora Aroyo and Davide Ceolin
March 28 2013
Social Web 2016 Lora Aroyo and Davide Ceolin
Social Web 2016 Lora Aroyo and Davide Ceolin
Social Web 2016 Lora Aroyo and Davide Ceolin
httpwwwwiredcommagazine201111mf_artsyall1
Social Web 2016 Lora Aroyo and Davide Ceolin
Collaborative Filteringbull Memory-based User-Item matrix ratingspreferences of users =gt
compute similarity between users amp recommend items of similar users
bull Model-based Item-Item matrix similarity (eg based on user ratings) between items =gt recommend items that are similar to the ones the user likes
bull Model-based Clustering cluster users according to their preferences =gt recommend items of users that belong to the same cluster
bull Model-based Bayesian networks P(u likes item B | u likes item A) = how likely is it that a user who likes item A will like item B learn probabilities from user ratingspreferences
bull Others rule-based other data mining techniques
Social Web 2016 Lora Aroyo and Davide Ceolin
bull complete input data is required
bull pre-computation not possible
bull does not scale well bull high quality of
recommendations
bull abstraction (model) of input data
bull pre-computation (partially) possible (model has to be re-built from time to time)
bull scales betterbull abstraction may reduce
recommendation quality
Memory vs Model-based
Social Web 2016 Lora Aroyo and Davide Ceolin
bull collaborative filtering lsquoneighborhoodsrsquo of people with similar interest amp recommending items based on likings in neighborhood
bull limitations next to lsquocold startrsquo and lsquosparsityrsquo the lack of control (over onersquos neighborhood) is also a problem ie cannot add lsquotrustedrsquo people nor exclude lsquostrangersquo ones
bull therefore interest in lsquosocial recommendersrsquo where presence of social connections defines the similarity in interests (eg social tagging CiteULike)bull does a social connection indicate user interest similaritybull how much users interest similarity depends on the strength
of their connectionbull is it feasible to use a social network as a personalized
recommendation[Lin amp Brusilovsky Social Networks and Interest Similarity The Case of CiteULike HTrsquo10]
Social Networks amp Interest Similarity
bull unilaterally connected pairs have more common itemsmetadatatags than non-connected pairsbull highest similarity for direct connections - decreasing with the increase of distance between users in SN bull reciprocal relationship users - significantly larger similarity than users in a unidirectional relationship bull traditional item-level similarity may be less reliable to find similar users in social bookmarking systemsbull peers connected by self-defined social connections could be a useful source for cross-recommendation
Conclusions
Social Web 2016 Lora Aroyo and Davide Ceolin
bull Input characteristics of items amp interests of a user into characteristics of items =gt Recommend items that feature characteristics which meet the userrsquos interests
bull Techniquesbull Data mining methods Cluster items based on their
characteristics =gt Infer usersrsquo interests into clustersbull IR methods Represent items amp users as term vectors
=gt Compute similarity between user profile vector and items
bull Utility-based methods Utility function that gets an item as input the parameters of the utility function are customized via preferences of a user
Content-based Recommendations
Social Web 2016 Lora Aroyo and Davide Ceolin
based on slides from Fabien Abel
Content Features
based on slides from Fabien Abel
User Model
based on slides from Fabien Abel
Recommendations
RecSys Issuesbull Cold-start problem (new user problem) nolittle data available to infer preferences of new
usersbull Changing User Preferences user interests may change over timebull Sparsity problem (new item problem) item descriptions are sparse eg not many user
rated or tagged an itembull Lack of Diversity (overfitting) when adapting too strongly to the preferences of users they
might see samesimilar recommendationsbull Use the right context users do things which might not be relevant for their user model
eg try out things do stuff for other peoplebull Research challenge right balance between serendipity amp personalizationbull Research challenge right way to use the influence of recommendations on userrsquos behavior
Social Web 2016 Lora Aroyo and Davide Ceolin
Announcementsbull Next deadlinesbull Tuesday March 1st 2359
Assignment 2bull Friday March 4th 1000 Post
questionbull Friday March 4th 1700 Vote
questions
image source httpwwwflickrcomphotosbionicteaching1375254387
Hands-on Teaser
bull Your Facebook Friendsrsquo popularity in a spread sheet
bull Locations of your Facebook Friendsbull Tag Cloud of your wall posts
- Social Web 2016
- Slide 2
- Slide 3
- Slide 4
- Slide 5
- Slide 6
- Slide 7
- Slide 8
- Slide 9
- Slide 10
- Slide 11
- Slide 12
- Slide 13
- Slide 14
- Slide 15
- Slide 16
- Slide 17
- Slide 18
- Slide 19
- Slide 20
- Slide 21
- Slide 22
- Slide 23
- Slide 24
- Slide 25
- Slide 26
- Slide 27
- Slide 28
- Slide 29
- Slide 30
- Slide 31
- Slide 32
- Slide 33
- Slide 34
- Slide 35
- Slide 36
- Slide 37
- Slide 38
- Slide 39
- Slide 40
- Slide 41
- Slide 42
- Slide 43
- Slide 44
- Slide 45
- Slide 46
- Slide 47
- Slide 48
- Announcements
- Slide 50
-
based on slides from Fabien Abel
Can we infer a Twitter-based User
Profile
User Modeling Building Blocks
based on slides from Fabien Abel
User Modeling Building Blocks
based on slides from Fabien Abel
User Modeling Building Blocks
based on slides from Fabien Abel
User Modeling Building Blocks
based on slides from Fabien Abel
Observationsbull Profile characteristics
bull Semantic enrichment solves sparsity problemsbull Profiles change over time recent profiles reflect
better current user demandsbull Temporal patterns weekend profiles differ
significantly from weekday profilesbull Impact on recommendations
bull The more fine-grained the concepts the better the recommendation performance entity-based gt topic-based gt hashtag-based
bull Semantic enrichment improves recommendation quality
bull Time-sensitivity (adapting to trends) improves performance
Social Web 2016 Lora Aroyo and Davide Ceolin
User Modelingit is not about putting everything in a user profile
it is about making the right choices
Social Web 2016 Lora Aroyo and Davide Ceolin
User Adaptation
Knowing the user to adapt a system or interfaceto improve the system functionality and user
experience
Social Web 2016 Lora Aroyo and Davide Ceolin
A Jameson Adaptive interfaces and agents The HCI handbook fundamentals evolving technologies and emerging applications pp 305ndash330 2003
User-Adaptive Systems
Social Web 2016 Lora Aroyo and Davide Ceolin
based on slides from Fabien Abel
Lastfm adapts to your music taste
bull Overfitting ldquobubble effectsrdquo loss of serendipity problem bull systems may adapt too strongly to the
interestsbehaviorbull eg an adaptive radio station may always play the
same or very similar songsbull We search for the right balance between novelty and
relevance for the userbull ldquoLost in Hyperspacerdquo problem
bull when adapting the navigation ndash ie the links on which users can click to findaccess information
bull eg re-orderinghiding of menu items may lead to confusion
Issues in User-Adaptive Systems
Social Web 2016 Lora Aroyo and Davide Ceolin
What is good user modelling amp
personalisation
httpwwwflickrcomphotosbellarosebyliz4729613108
From the consumer perspective of an adaptive system
From the provider perspective of an adaptive system
Success Perspectives
Social Web 2016 Lora Aroyo and Davide Ceolin
bull User studies askobserve (selected) people whether you did a good job
bull Log analysis Analyze (click) data and infer whether you did a good job
bull Evaluation of user modelingbull measure quality of profiles directly eg measure
overlap with existing (true) profiles or let people judge the quality of the generated user profiles
bull measure quality of application that exploits the user profile eg apply user modeling strategies in a recommender system
Evaluation Strategies
Social Web 2016 Lora Aroyo and Davide Ceolin
Evaluating User Modeling in RecSys
Social Web 2016 Lora Aroyo and Davide Ceolin
Possible Metrics
Social Web 2016 Lora Aroyo and Davide Ceolin
bull The usual IR metricsbull Precision fraction of retrieved items that are relevantbull Recall fraction of relevant items that have been
retrievedbull F-Measure (harmonic) mean of precision and recall
bull Metrics for evaluating recommendation (rankings)bull Mean Reciprocal Rank (MRR) of first relevant itembull Successk probability that relevant item occurs
within the top kbull If a true ranking is given rank correlations bull Precisionk Recallk amp F-Measurek
bull Metrics for evaluating prediction of user preferencesbull MAE = Mean Absolute Errorbull TrueFalse PositivesNegatives
Predict relevantusefulinteresting items for a given user (in a given context) itrsquos often a ranking task
Recommendation Systems
Social Web 2016 Lora Aroyo and Davide Ceolin
March 28 2013
Social Web 2016 Lora Aroyo and Davide Ceolin
Social Web 2016 Lora Aroyo and Davide Ceolin
Social Web 2016 Lora Aroyo and Davide Ceolin
httpwwwwiredcommagazine201111mf_artsyall1
Social Web 2016 Lora Aroyo and Davide Ceolin
Collaborative Filteringbull Memory-based User-Item matrix ratingspreferences of users =gt
compute similarity between users amp recommend items of similar users
bull Model-based Item-Item matrix similarity (eg based on user ratings) between items =gt recommend items that are similar to the ones the user likes
bull Model-based Clustering cluster users according to their preferences =gt recommend items of users that belong to the same cluster
bull Model-based Bayesian networks P(u likes item B | u likes item A) = how likely is it that a user who likes item A will like item B learn probabilities from user ratingspreferences
bull Others rule-based other data mining techniques
Social Web 2016 Lora Aroyo and Davide Ceolin
bull complete input data is required
bull pre-computation not possible
bull does not scale well bull high quality of
recommendations
bull abstraction (model) of input data
bull pre-computation (partially) possible (model has to be re-built from time to time)
bull scales betterbull abstraction may reduce
recommendation quality
Memory vs Model-based
Social Web 2016 Lora Aroyo and Davide Ceolin
bull collaborative filtering lsquoneighborhoodsrsquo of people with similar interest amp recommending items based on likings in neighborhood
bull limitations next to lsquocold startrsquo and lsquosparsityrsquo the lack of control (over onersquos neighborhood) is also a problem ie cannot add lsquotrustedrsquo people nor exclude lsquostrangersquo ones
bull therefore interest in lsquosocial recommendersrsquo where presence of social connections defines the similarity in interests (eg social tagging CiteULike)bull does a social connection indicate user interest similaritybull how much users interest similarity depends on the strength
of their connectionbull is it feasible to use a social network as a personalized
recommendation[Lin amp Brusilovsky Social Networks and Interest Similarity The Case of CiteULike HTrsquo10]
Social Networks amp Interest Similarity
bull unilaterally connected pairs have more common itemsmetadatatags than non-connected pairsbull highest similarity for direct connections - decreasing with the increase of distance between users in SN bull reciprocal relationship users - significantly larger similarity than users in a unidirectional relationship bull traditional item-level similarity may be less reliable to find similar users in social bookmarking systemsbull peers connected by self-defined social connections could be a useful source for cross-recommendation
Conclusions
Social Web 2016 Lora Aroyo and Davide Ceolin
bull Input characteristics of items amp interests of a user into characteristics of items =gt Recommend items that feature characteristics which meet the userrsquos interests
bull Techniquesbull Data mining methods Cluster items based on their
characteristics =gt Infer usersrsquo interests into clustersbull IR methods Represent items amp users as term vectors
=gt Compute similarity between user profile vector and items
bull Utility-based methods Utility function that gets an item as input the parameters of the utility function are customized via preferences of a user
Content-based Recommendations
Social Web 2016 Lora Aroyo and Davide Ceolin
based on slides from Fabien Abel
Content Features
based on slides from Fabien Abel
User Model
based on slides from Fabien Abel
Recommendations
RecSys Issuesbull Cold-start problem (new user problem) nolittle data available to infer preferences of new
usersbull Changing User Preferences user interests may change over timebull Sparsity problem (new item problem) item descriptions are sparse eg not many user
rated or tagged an itembull Lack of Diversity (overfitting) when adapting too strongly to the preferences of users they
might see samesimilar recommendationsbull Use the right context users do things which might not be relevant for their user model
eg try out things do stuff for other peoplebull Research challenge right balance between serendipity amp personalizationbull Research challenge right way to use the influence of recommendations on userrsquos behavior
Social Web 2016 Lora Aroyo and Davide Ceolin
Announcementsbull Next deadlinesbull Tuesday March 1st 2359
Assignment 2bull Friday March 4th 1000 Post
questionbull Friday March 4th 1700 Vote
questions
image source httpwwwflickrcomphotosbionicteaching1375254387
Hands-on Teaser
bull Your Facebook Friendsrsquo popularity in a spread sheet
bull Locations of your Facebook Friendsbull Tag Cloud of your wall posts
- Social Web 2016
- Slide 2
- Slide 3
- Slide 4
- Slide 5
- Slide 6
- Slide 7
- Slide 8
- Slide 9
- Slide 10
- Slide 11
- Slide 12
- Slide 13
- Slide 14
- Slide 15
- Slide 16
- Slide 17
- Slide 18
- Slide 19
- Slide 20
- Slide 21
- Slide 22
- Slide 23
- Slide 24
- Slide 25
- Slide 26
- Slide 27
- Slide 28
- Slide 29
- Slide 30
- Slide 31
- Slide 32
- Slide 33
- Slide 34
- Slide 35
- Slide 36
- Slide 37
- Slide 38
- Slide 39
- Slide 40
- Slide 41
- Slide 42
- Slide 43
- Slide 44
- Slide 45
- Slide 46
- Slide 47
- Slide 48
- Announcements
- Slide 50
-
User Modeling Building Blocks
based on slides from Fabien Abel
User Modeling Building Blocks
based on slides from Fabien Abel
User Modeling Building Blocks
based on slides from Fabien Abel
User Modeling Building Blocks
based on slides from Fabien Abel
Observationsbull Profile characteristics
bull Semantic enrichment solves sparsity problemsbull Profiles change over time recent profiles reflect
better current user demandsbull Temporal patterns weekend profiles differ
significantly from weekday profilesbull Impact on recommendations
bull The more fine-grained the concepts the better the recommendation performance entity-based gt topic-based gt hashtag-based
bull Semantic enrichment improves recommendation quality
bull Time-sensitivity (adapting to trends) improves performance
Social Web 2016 Lora Aroyo and Davide Ceolin
User Modelingit is not about putting everything in a user profile
it is about making the right choices
Social Web 2016 Lora Aroyo and Davide Ceolin
User Adaptation
Knowing the user to adapt a system or interfaceto improve the system functionality and user
experience
Social Web 2016 Lora Aroyo and Davide Ceolin
A Jameson Adaptive interfaces and agents The HCI handbook fundamentals evolving technologies and emerging applications pp 305ndash330 2003
User-Adaptive Systems
Social Web 2016 Lora Aroyo and Davide Ceolin
based on slides from Fabien Abel
Lastfm adapts to your music taste
bull Overfitting ldquobubble effectsrdquo loss of serendipity problem bull systems may adapt too strongly to the
interestsbehaviorbull eg an adaptive radio station may always play the
same or very similar songsbull We search for the right balance between novelty and
relevance for the userbull ldquoLost in Hyperspacerdquo problem
bull when adapting the navigation ndash ie the links on which users can click to findaccess information
bull eg re-orderinghiding of menu items may lead to confusion
Issues in User-Adaptive Systems
Social Web 2016 Lora Aroyo and Davide Ceolin
What is good user modelling amp
personalisation
httpwwwflickrcomphotosbellarosebyliz4729613108
From the consumer perspective of an adaptive system
From the provider perspective of an adaptive system
Success Perspectives
Social Web 2016 Lora Aroyo and Davide Ceolin
bull User studies askobserve (selected) people whether you did a good job
bull Log analysis Analyze (click) data and infer whether you did a good job
bull Evaluation of user modelingbull measure quality of profiles directly eg measure
overlap with existing (true) profiles or let people judge the quality of the generated user profiles
bull measure quality of application that exploits the user profile eg apply user modeling strategies in a recommender system
Evaluation Strategies
Social Web 2016 Lora Aroyo and Davide Ceolin
Evaluating User Modeling in RecSys
Social Web 2016 Lora Aroyo and Davide Ceolin
Possible Metrics
Social Web 2016 Lora Aroyo and Davide Ceolin
bull The usual IR metricsbull Precision fraction of retrieved items that are relevantbull Recall fraction of relevant items that have been
retrievedbull F-Measure (harmonic) mean of precision and recall
bull Metrics for evaluating recommendation (rankings)bull Mean Reciprocal Rank (MRR) of first relevant itembull Successk probability that relevant item occurs
within the top kbull If a true ranking is given rank correlations bull Precisionk Recallk amp F-Measurek
bull Metrics for evaluating prediction of user preferencesbull MAE = Mean Absolute Errorbull TrueFalse PositivesNegatives
Predict relevantusefulinteresting items for a given user (in a given context) itrsquos often a ranking task
Recommendation Systems
Social Web 2016 Lora Aroyo and Davide Ceolin
March 28 2013
Social Web 2016 Lora Aroyo and Davide Ceolin
Social Web 2016 Lora Aroyo and Davide Ceolin
Social Web 2016 Lora Aroyo and Davide Ceolin
httpwwwwiredcommagazine201111mf_artsyall1
Social Web 2016 Lora Aroyo and Davide Ceolin
Collaborative Filteringbull Memory-based User-Item matrix ratingspreferences of users =gt
compute similarity between users amp recommend items of similar users
bull Model-based Item-Item matrix similarity (eg based on user ratings) between items =gt recommend items that are similar to the ones the user likes
bull Model-based Clustering cluster users according to their preferences =gt recommend items of users that belong to the same cluster
bull Model-based Bayesian networks P(u likes item B | u likes item A) = how likely is it that a user who likes item A will like item B learn probabilities from user ratingspreferences
bull Others rule-based other data mining techniques
Social Web 2016 Lora Aroyo and Davide Ceolin
bull complete input data is required
bull pre-computation not possible
bull does not scale well bull high quality of
recommendations
bull abstraction (model) of input data
bull pre-computation (partially) possible (model has to be re-built from time to time)
bull scales betterbull abstraction may reduce
recommendation quality
Memory vs Model-based
Social Web 2016 Lora Aroyo and Davide Ceolin
bull collaborative filtering lsquoneighborhoodsrsquo of people with similar interest amp recommending items based on likings in neighborhood
bull limitations next to lsquocold startrsquo and lsquosparsityrsquo the lack of control (over onersquos neighborhood) is also a problem ie cannot add lsquotrustedrsquo people nor exclude lsquostrangersquo ones
bull therefore interest in lsquosocial recommendersrsquo where presence of social connections defines the similarity in interests (eg social tagging CiteULike)bull does a social connection indicate user interest similaritybull how much users interest similarity depends on the strength
of their connectionbull is it feasible to use a social network as a personalized
recommendation[Lin amp Brusilovsky Social Networks and Interest Similarity The Case of CiteULike HTrsquo10]
Social Networks amp Interest Similarity
bull unilaterally connected pairs have more common itemsmetadatatags than non-connected pairsbull highest similarity for direct connections - decreasing with the increase of distance between users in SN bull reciprocal relationship users - significantly larger similarity than users in a unidirectional relationship bull traditional item-level similarity may be less reliable to find similar users in social bookmarking systemsbull peers connected by self-defined social connections could be a useful source for cross-recommendation
Conclusions
Social Web 2016 Lora Aroyo and Davide Ceolin
bull Input characteristics of items amp interests of a user into characteristics of items =gt Recommend items that feature characteristics which meet the userrsquos interests
bull Techniquesbull Data mining methods Cluster items based on their
characteristics =gt Infer usersrsquo interests into clustersbull IR methods Represent items amp users as term vectors
=gt Compute similarity between user profile vector and items
bull Utility-based methods Utility function that gets an item as input the parameters of the utility function are customized via preferences of a user
Content-based Recommendations
Social Web 2016 Lora Aroyo and Davide Ceolin
based on slides from Fabien Abel
Content Features
based on slides from Fabien Abel
User Model
based on slides from Fabien Abel
Recommendations
RecSys Issuesbull Cold-start problem (new user problem) nolittle data available to infer preferences of new
usersbull Changing User Preferences user interests may change over timebull Sparsity problem (new item problem) item descriptions are sparse eg not many user
rated or tagged an itembull Lack of Diversity (overfitting) when adapting too strongly to the preferences of users they
might see samesimilar recommendationsbull Use the right context users do things which might not be relevant for their user model
eg try out things do stuff for other peoplebull Research challenge right balance between serendipity amp personalizationbull Research challenge right way to use the influence of recommendations on userrsquos behavior
Social Web 2016 Lora Aroyo and Davide Ceolin
Announcementsbull Next deadlinesbull Tuesday March 1st 2359
Assignment 2bull Friday March 4th 1000 Post
questionbull Friday March 4th 1700 Vote
questions
image source httpwwwflickrcomphotosbionicteaching1375254387
Hands-on Teaser
bull Your Facebook Friendsrsquo popularity in a spread sheet
bull Locations of your Facebook Friendsbull Tag Cloud of your wall posts
- Social Web 2016
- Slide 2
- Slide 3
- Slide 4
- Slide 5
- Slide 6
- Slide 7
- Slide 8
- Slide 9
- Slide 10
- Slide 11
- Slide 12
- Slide 13
- Slide 14
- Slide 15
- Slide 16
- Slide 17
- Slide 18
- Slide 19
- Slide 20
- Slide 21
- Slide 22
- Slide 23
- Slide 24
- Slide 25
- Slide 26
- Slide 27
- Slide 28
- Slide 29
- Slide 30
- Slide 31
- Slide 32
- Slide 33
- Slide 34
- Slide 35
- Slide 36
- Slide 37
- Slide 38
- Slide 39
- Slide 40
- Slide 41
- Slide 42
- Slide 43
- Slide 44
- Slide 45
- Slide 46
- Slide 47
- Slide 48
- Announcements
- Slide 50
-
User Modeling Building Blocks
based on slides from Fabien Abel
User Modeling Building Blocks
based on slides from Fabien Abel
User Modeling Building Blocks
based on slides from Fabien Abel
Observationsbull Profile characteristics
bull Semantic enrichment solves sparsity problemsbull Profiles change over time recent profiles reflect
better current user demandsbull Temporal patterns weekend profiles differ
significantly from weekday profilesbull Impact on recommendations
bull The more fine-grained the concepts the better the recommendation performance entity-based gt topic-based gt hashtag-based
bull Semantic enrichment improves recommendation quality
bull Time-sensitivity (adapting to trends) improves performance
Social Web 2016 Lora Aroyo and Davide Ceolin
User Modelingit is not about putting everything in a user profile
it is about making the right choices
Social Web 2016 Lora Aroyo and Davide Ceolin
User Adaptation
Knowing the user to adapt a system or interfaceto improve the system functionality and user
experience
Social Web 2016 Lora Aroyo and Davide Ceolin
A Jameson Adaptive interfaces and agents The HCI handbook fundamentals evolving technologies and emerging applications pp 305ndash330 2003
User-Adaptive Systems
Social Web 2016 Lora Aroyo and Davide Ceolin
based on slides from Fabien Abel
Lastfm adapts to your music taste
bull Overfitting ldquobubble effectsrdquo loss of serendipity problem bull systems may adapt too strongly to the
interestsbehaviorbull eg an adaptive radio station may always play the
same or very similar songsbull We search for the right balance between novelty and
relevance for the userbull ldquoLost in Hyperspacerdquo problem
bull when adapting the navigation ndash ie the links on which users can click to findaccess information
bull eg re-orderinghiding of menu items may lead to confusion
Issues in User-Adaptive Systems
Social Web 2016 Lora Aroyo and Davide Ceolin
What is good user modelling amp
personalisation
httpwwwflickrcomphotosbellarosebyliz4729613108
From the consumer perspective of an adaptive system
From the provider perspective of an adaptive system
Success Perspectives
Social Web 2016 Lora Aroyo and Davide Ceolin
bull User studies askobserve (selected) people whether you did a good job
bull Log analysis Analyze (click) data and infer whether you did a good job
bull Evaluation of user modelingbull measure quality of profiles directly eg measure
overlap with existing (true) profiles or let people judge the quality of the generated user profiles
bull measure quality of application that exploits the user profile eg apply user modeling strategies in a recommender system
Evaluation Strategies
Social Web 2016 Lora Aroyo and Davide Ceolin
Evaluating User Modeling in RecSys
Social Web 2016 Lora Aroyo and Davide Ceolin
Possible Metrics
Social Web 2016 Lora Aroyo and Davide Ceolin
bull The usual IR metricsbull Precision fraction of retrieved items that are relevantbull Recall fraction of relevant items that have been
retrievedbull F-Measure (harmonic) mean of precision and recall
bull Metrics for evaluating recommendation (rankings)bull Mean Reciprocal Rank (MRR) of first relevant itembull Successk probability that relevant item occurs
within the top kbull If a true ranking is given rank correlations bull Precisionk Recallk amp F-Measurek
bull Metrics for evaluating prediction of user preferencesbull MAE = Mean Absolute Errorbull TrueFalse PositivesNegatives
Predict relevantusefulinteresting items for a given user (in a given context) itrsquos often a ranking task
Recommendation Systems
Social Web 2016 Lora Aroyo and Davide Ceolin
March 28 2013
Social Web 2016 Lora Aroyo and Davide Ceolin
Social Web 2016 Lora Aroyo and Davide Ceolin
Social Web 2016 Lora Aroyo and Davide Ceolin
httpwwwwiredcommagazine201111mf_artsyall1
Social Web 2016 Lora Aroyo and Davide Ceolin
Collaborative Filteringbull Memory-based User-Item matrix ratingspreferences of users =gt
compute similarity between users amp recommend items of similar users
bull Model-based Item-Item matrix similarity (eg based on user ratings) between items =gt recommend items that are similar to the ones the user likes
bull Model-based Clustering cluster users according to their preferences =gt recommend items of users that belong to the same cluster
bull Model-based Bayesian networks P(u likes item B | u likes item A) = how likely is it that a user who likes item A will like item B learn probabilities from user ratingspreferences
bull Others rule-based other data mining techniques
Social Web 2016 Lora Aroyo and Davide Ceolin
bull complete input data is required
bull pre-computation not possible
bull does not scale well bull high quality of
recommendations
bull abstraction (model) of input data
bull pre-computation (partially) possible (model has to be re-built from time to time)
bull scales betterbull abstraction may reduce
recommendation quality
Memory vs Model-based
Social Web 2016 Lora Aroyo and Davide Ceolin
bull collaborative filtering lsquoneighborhoodsrsquo of people with similar interest amp recommending items based on likings in neighborhood
bull limitations next to lsquocold startrsquo and lsquosparsityrsquo the lack of control (over onersquos neighborhood) is also a problem ie cannot add lsquotrustedrsquo people nor exclude lsquostrangersquo ones
bull therefore interest in lsquosocial recommendersrsquo where presence of social connections defines the similarity in interests (eg social tagging CiteULike)bull does a social connection indicate user interest similaritybull how much users interest similarity depends on the strength
of their connectionbull is it feasible to use a social network as a personalized
recommendation[Lin amp Brusilovsky Social Networks and Interest Similarity The Case of CiteULike HTrsquo10]
Social Networks amp Interest Similarity
bull unilaterally connected pairs have more common itemsmetadatatags than non-connected pairsbull highest similarity for direct connections - decreasing with the increase of distance between users in SN bull reciprocal relationship users - significantly larger similarity than users in a unidirectional relationship bull traditional item-level similarity may be less reliable to find similar users in social bookmarking systemsbull peers connected by self-defined social connections could be a useful source for cross-recommendation
Conclusions
Social Web 2016 Lora Aroyo and Davide Ceolin
bull Input characteristics of items amp interests of a user into characteristics of items =gt Recommend items that feature characteristics which meet the userrsquos interests
bull Techniquesbull Data mining methods Cluster items based on their
characteristics =gt Infer usersrsquo interests into clustersbull IR methods Represent items amp users as term vectors
=gt Compute similarity between user profile vector and items
bull Utility-based methods Utility function that gets an item as input the parameters of the utility function are customized via preferences of a user
Content-based Recommendations
Social Web 2016 Lora Aroyo and Davide Ceolin
based on slides from Fabien Abel
Content Features
based on slides from Fabien Abel
User Model
based on slides from Fabien Abel
Recommendations
RecSys Issuesbull Cold-start problem (new user problem) nolittle data available to infer preferences of new
usersbull Changing User Preferences user interests may change over timebull Sparsity problem (new item problem) item descriptions are sparse eg not many user
rated or tagged an itembull Lack of Diversity (overfitting) when adapting too strongly to the preferences of users they
might see samesimilar recommendationsbull Use the right context users do things which might not be relevant for their user model
eg try out things do stuff for other peoplebull Research challenge right balance between serendipity amp personalizationbull Research challenge right way to use the influence of recommendations on userrsquos behavior
Social Web 2016 Lora Aroyo and Davide Ceolin
Announcementsbull Next deadlinesbull Tuesday March 1st 2359
Assignment 2bull Friday March 4th 1000 Post
questionbull Friday March 4th 1700 Vote
questions
image source httpwwwflickrcomphotosbionicteaching1375254387
Hands-on Teaser
bull Your Facebook Friendsrsquo popularity in a spread sheet
bull Locations of your Facebook Friendsbull Tag Cloud of your wall posts
- Social Web 2016
- Slide 2
- Slide 3
- Slide 4
- Slide 5
- Slide 6
- Slide 7
- Slide 8
- Slide 9
- Slide 10
- Slide 11
- Slide 12
- Slide 13
- Slide 14
- Slide 15
- Slide 16
- Slide 17
- Slide 18
- Slide 19
- Slide 20
- Slide 21
- Slide 22
- Slide 23
- Slide 24
- Slide 25
- Slide 26
- Slide 27
- Slide 28
- Slide 29
- Slide 30
- Slide 31
- Slide 32
- Slide 33
- Slide 34
- Slide 35
- Slide 36
- Slide 37
- Slide 38
- Slide 39
- Slide 40
- Slide 41
- Slide 42
- Slide 43
- Slide 44
- Slide 45
- Slide 46
- Slide 47
- Slide 48
- Announcements
- Slide 50
-
User Modeling Building Blocks
based on slides from Fabien Abel
User Modeling Building Blocks
based on slides from Fabien Abel
Observationsbull Profile characteristics
bull Semantic enrichment solves sparsity problemsbull Profiles change over time recent profiles reflect
better current user demandsbull Temporal patterns weekend profiles differ
significantly from weekday profilesbull Impact on recommendations
bull The more fine-grained the concepts the better the recommendation performance entity-based gt topic-based gt hashtag-based
bull Semantic enrichment improves recommendation quality
bull Time-sensitivity (adapting to trends) improves performance
Social Web 2016 Lora Aroyo and Davide Ceolin
User Modelingit is not about putting everything in a user profile
it is about making the right choices
Social Web 2016 Lora Aroyo and Davide Ceolin
User Adaptation
Knowing the user to adapt a system or interfaceto improve the system functionality and user
experience
Social Web 2016 Lora Aroyo and Davide Ceolin
A Jameson Adaptive interfaces and agents The HCI handbook fundamentals evolving technologies and emerging applications pp 305ndash330 2003
User-Adaptive Systems
Social Web 2016 Lora Aroyo and Davide Ceolin
based on slides from Fabien Abel
Lastfm adapts to your music taste
bull Overfitting ldquobubble effectsrdquo loss of serendipity problem bull systems may adapt too strongly to the
interestsbehaviorbull eg an adaptive radio station may always play the
same or very similar songsbull We search for the right balance between novelty and
relevance for the userbull ldquoLost in Hyperspacerdquo problem
bull when adapting the navigation ndash ie the links on which users can click to findaccess information
bull eg re-orderinghiding of menu items may lead to confusion
Issues in User-Adaptive Systems
Social Web 2016 Lora Aroyo and Davide Ceolin
What is good user modelling amp
personalisation
httpwwwflickrcomphotosbellarosebyliz4729613108
From the consumer perspective of an adaptive system
From the provider perspective of an adaptive system
Success Perspectives
Social Web 2016 Lora Aroyo and Davide Ceolin
bull User studies askobserve (selected) people whether you did a good job
bull Log analysis Analyze (click) data and infer whether you did a good job
bull Evaluation of user modelingbull measure quality of profiles directly eg measure
overlap with existing (true) profiles or let people judge the quality of the generated user profiles
bull measure quality of application that exploits the user profile eg apply user modeling strategies in a recommender system
Evaluation Strategies
Social Web 2016 Lora Aroyo and Davide Ceolin
Evaluating User Modeling in RecSys
Social Web 2016 Lora Aroyo and Davide Ceolin
Possible Metrics
Social Web 2016 Lora Aroyo and Davide Ceolin
bull The usual IR metricsbull Precision fraction of retrieved items that are relevantbull Recall fraction of relevant items that have been
retrievedbull F-Measure (harmonic) mean of precision and recall
bull Metrics for evaluating recommendation (rankings)bull Mean Reciprocal Rank (MRR) of first relevant itembull Successk probability that relevant item occurs
within the top kbull If a true ranking is given rank correlations bull Precisionk Recallk amp F-Measurek
bull Metrics for evaluating prediction of user preferencesbull MAE = Mean Absolute Errorbull TrueFalse PositivesNegatives
Predict relevantusefulinteresting items for a given user (in a given context) itrsquos often a ranking task
Recommendation Systems
Social Web 2016 Lora Aroyo and Davide Ceolin
March 28 2013
Social Web 2016 Lora Aroyo and Davide Ceolin
Social Web 2016 Lora Aroyo and Davide Ceolin
Social Web 2016 Lora Aroyo and Davide Ceolin
httpwwwwiredcommagazine201111mf_artsyall1
Social Web 2016 Lora Aroyo and Davide Ceolin
Collaborative Filteringbull Memory-based User-Item matrix ratingspreferences of users =gt
compute similarity between users amp recommend items of similar users
bull Model-based Item-Item matrix similarity (eg based on user ratings) between items =gt recommend items that are similar to the ones the user likes
bull Model-based Clustering cluster users according to their preferences =gt recommend items of users that belong to the same cluster
bull Model-based Bayesian networks P(u likes item B | u likes item A) = how likely is it that a user who likes item A will like item B learn probabilities from user ratingspreferences
bull Others rule-based other data mining techniques
Social Web 2016 Lora Aroyo and Davide Ceolin
bull complete input data is required
bull pre-computation not possible
bull does not scale well bull high quality of
recommendations
bull abstraction (model) of input data
bull pre-computation (partially) possible (model has to be re-built from time to time)
bull scales betterbull abstraction may reduce
recommendation quality
Memory vs Model-based
Social Web 2016 Lora Aroyo and Davide Ceolin
bull collaborative filtering lsquoneighborhoodsrsquo of people with similar interest amp recommending items based on likings in neighborhood
bull limitations next to lsquocold startrsquo and lsquosparsityrsquo the lack of control (over onersquos neighborhood) is also a problem ie cannot add lsquotrustedrsquo people nor exclude lsquostrangersquo ones
bull therefore interest in lsquosocial recommendersrsquo where presence of social connections defines the similarity in interests (eg social tagging CiteULike)bull does a social connection indicate user interest similaritybull how much users interest similarity depends on the strength
of their connectionbull is it feasible to use a social network as a personalized
recommendation[Lin amp Brusilovsky Social Networks and Interest Similarity The Case of CiteULike HTrsquo10]
Social Networks amp Interest Similarity
bull unilaterally connected pairs have more common itemsmetadatatags than non-connected pairsbull highest similarity for direct connections - decreasing with the increase of distance between users in SN bull reciprocal relationship users - significantly larger similarity than users in a unidirectional relationship bull traditional item-level similarity may be less reliable to find similar users in social bookmarking systemsbull peers connected by self-defined social connections could be a useful source for cross-recommendation
Conclusions
Social Web 2016 Lora Aroyo and Davide Ceolin
bull Input characteristics of items amp interests of a user into characteristics of items =gt Recommend items that feature characteristics which meet the userrsquos interests
bull Techniquesbull Data mining methods Cluster items based on their
characteristics =gt Infer usersrsquo interests into clustersbull IR methods Represent items amp users as term vectors
=gt Compute similarity between user profile vector and items
bull Utility-based methods Utility function that gets an item as input the parameters of the utility function are customized via preferences of a user
Content-based Recommendations
Social Web 2016 Lora Aroyo and Davide Ceolin
based on slides from Fabien Abel
Content Features
based on slides from Fabien Abel
User Model
based on slides from Fabien Abel
Recommendations
RecSys Issuesbull Cold-start problem (new user problem) nolittle data available to infer preferences of new
usersbull Changing User Preferences user interests may change over timebull Sparsity problem (new item problem) item descriptions are sparse eg not many user
rated or tagged an itembull Lack of Diversity (overfitting) when adapting too strongly to the preferences of users they
might see samesimilar recommendationsbull Use the right context users do things which might not be relevant for their user model
eg try out things do stuff for other peoplebull Research challenge right balance between serendipity amp personalizationbull Research challenge right way to use the influence of recommendations on userrsquos behavior
Social Web 2016 Lora Aroyo and Davide Ceolin
Announcementsbull Next deadlinesbull Tuesday March 1st 2359
Assignment 2bull Friday March 4th 1000 Post
questionbull Friday March 4th 1700 Vote
questions
image source httpwwwflickrcomphotosbionicteaching1375254387
Hands-on Teaser
bull Your Facebook Friendsrsquo popularity in a spread sheet
bull Locations of your Facebook Friendsbull Tag Cloud of your wall posts
- Social Web 2016
- Slide 2
- Slide 3
- Slide 4
- Slide 5
- Slide 6
- Slide 7
- Slide 8
- Slide 9
- Slide 10
- Slide 11
- Slide 12
- Slide 13
- Slide 14
- Slide 15
- Slide 16
- Slide 17
- Slide 18
- Slide 19
- Slide 20
- Slide 21
- Slide 22
- Slide 23
- Slide 24
- Slide 25
- Slide 26
- Slide 27
- Slide 28
- Slide 29
- Slide 30
- Slide 31
- Slide 32
- Slide 33
- Slide 34
- Slide 35
- Slide 36
- Slide 37
- Slide 38
- Slide 39
- Slide 40
- Slide 41
- Slide 42
- Slide 43
- Slide 44
- Slide 45
- Slide 46
- Slide 47
- Slide 48
- Announcements
- Slide 50
-
User Modeling Building Blocks
based on slides from Fabien Abel
Observationsbull Profile characteristics
bull Semantic enrichment solves sparsity problemsbull Profiles change over time recent profiles reflect
better current user demandsbull Temporal patterns weekend profiles differ
significantly from weekday profilesbull Impact on recommendations
bull The more fine-grained the concepts the better the recommendation performance entity-based gt topic-based gt hashtag-based
bull Semantic enrichment improves recommendation quality
bull Time-sensitivity (adapting to trends) improves performance
Social Web 2016 Lora Aroyo and Davide Ceolin
User Modelingit is not about putting everything in a user profile
it is about making the right choices
Social Web 2016 Lora Aroyo and Davide Ceolin
User Adaptation
Knowing the user to adapt a system or interfaceto improve the system functionality and user
experience
Social Web 2016 Lora Aroyo and Davide Ceolin
A Jameson Adaptive interfaces and agents The HCI handbook fundamentals evolving technologies and emerging applications pp 305ndash330 2003
User-Adaptive Systems
Social Web 2016 Lora Aroyo and Davide Ceolin
based on slides from Fabien Abel
Lastfm adapts to your music taste
bull Overfitting ldquobubble effectsrdquo loss of serendipity problem bull systems may adapt too strongly to the
interestsbehaviorbull eg an adaptive radio station may always play the
same or very similar songsbull We search for the right balance between novelty and
relevance for the userbull ldquoLost in Hyperspacerdquo problem
bull when adapting the navigation ndash ie the links on which users can click to findaccess information
bull eg re-orderinghiding of menu items may lead to confusion
Issues in User-Adaptive Systems
Social Web 2016 Lora Aroyo and Davide Ceolin
What is good user modelling amp
personalisation
httpwwwflickrcomphotosbellarosebyliz4729613108
From the consumer perspective of an adaptive system
From the provider perspective of an adaptive system
Success Perspectives
Social Web 2016 Lora Aroyo and Davide Ceolin
bull User studies askobserve (selected) people whether you did a good job
bull Log analysis Analyze (click) data and infer whether you did a good job
bull Evaluation of user modelingbull measure quality of profiles directly eg measure
overlap with existing (true) profiles or let people judge the quality of the generated user profiles
bull measure quality of application that exploits the user profile eg apply user modeling strategies in a recommender system
Evaluation Strategies
Social Web 2016 Lora Aroyo and Davide Ceolin
Evaluating User Modeling in RecSys
Social Web 2016 Lora Aroyo and Davide Ceolin
Possible Metrics
Social Web 2016 Lora Aroyo and Davide Ceolin
bull The usual IR metricsbull Precision fraction of retrieved items that are relevantbull Recall fraction of relevant items that have been
retrievedbull F-Measure (harmonic) mean of precision and recall
bull Metrics for evaluating recommendation (rankings)bull Mean Reciprocal Rank (MRR) of first relevant itembull Successk probability that relevant item occurs
within the top kbull If a true ranking is given rank correlations bull Precisionk Recallk amp F-Measurek
bull Metrics for evaluating prediction of user preferencesbull MAE = Mean Absolute Errorbull TrueFalse PositivesNegatives
Predict relevantusefulinteresting items for a given user (in a given context) itrsquos often a ranking task
Recommendation Systems
Social Web 2016 Lora Aroyo and Davide Ceolin
March 28 2013
Social Web 2016 Lora Aroyo and Davide Ceolin
Social Web 2016 Lora Aroyo and Davide Ceolin
Social Web 2016 Lora Aroyo and Davide Ceolin
httpwwwwiredcommagazine201111mf_artsyall1
Social Web 2016 Lora Aroyo and Davide Ceolin
Collaborative Filteringbull Memory-based User-Item matrix ratingspreferences of users =gt
compute similarity between users amp recommend items of similar users
bull Model-based Item-Item matrix similarity (eg based on user ratings) between items =gt recommend items that are similar to the ones the user likes
bull Model-based Clustering cluster users according to their preferences =gt recommend items of users that belong to the same cluster
bull Model-based Bayesian networks P(u likes item B | u likes item A) = how likely is it that a user who likes item A will like item B learn probabilities from user ratingspreferences
bull Others rule-based other data mining techniques
Social Web 2016 Lora Aroyo and Davide Ceolin
bull complete input data is required
bull pre-computation not possible
bull does not scale well bull high quality of
recommendations
bull abstraction (model) of input data
bull pre-computation (partially) possible (model has to be re-built from time to time)
bull scales betterbull abstraction may reduce
recommendation quality
Memory vs Model-based
Social Web 2016 Lora Aroyo and Davide Ceolin
bull collaborative filtering lsquoneighborhoodsrsquo of people with similar interest amp recommending items based on likings in neighborhood
bull limitations next to lsquocold startrsquo and lsquosparsityrsquo the lack of control (over onersquos neighborhood) is also a problem ie cannot add lsquotrustedrsquo people nor exclude lsquostrangersquo ones
bull therefore interest in lsquosocial recommendersrsquo where presence of social connections defines the similarity in interests (eg social tagging CiteULike)bull does a social connection indicate user interest similaritybull how much users interest similarity depends on the strength
of their connectionbull is it feasible to use a social network as a personalized
recommendation[Lin amp Brusilovsky Social Networks and Interest Similarity The Case of CiteULike HTrsquo10]
Social Networks amp Interest Similarity
bull unilaterally connected pairs have more common itemsmetadatatags than non-connected pairsbull highest similarity for direct connections - decreasing with the increase of distance between users in SN bull reciprocal relationship users - significantly larger similarity than users in a unidirectional relationship bull traditional item-level similarity may be less reliable to find similar users in social bookmarking systemsbull peers connected by self-defined social connections could be a useful source for cross-recommendation
Conclusions
Social Web 2016 Lora Aroyo and Davide Ceolin
bull Input characteristics of items amp interests of a user into characteristics of items =gt Recommend items that feature characteristics which meet the userrsquos interests
bull Techniquesbull Data mining methods Cluster items based on their
characteristics =gt Infer usersrsquo interests into clustersbull IR methods Represent items amp users as term vectors
=gt Compute similarity between user profile vector and items
bull Utility-based methods Utility function that gets an item as input the parameters of the utility function are customized via preferences of a user
Content-based Recommendations
Social Web 2016 Lora Aroyo and Davide Ceolin
based on slides from Fabien Abel
Content Features
based on slides from Fabien Abel
User Model
based on slides from Fabien Abel
Recommendations
RecSys Issuesbull Cold-start problem (new user problem) nolittle data available to infer preferences of new
usersbull Changing User Preferences user interests may change over timebull Sparsity problem (new item problem) item descriptions are sparse eg not many user
rated or tagged an itembull Lack of Diversity (overfitting) when adapting too strongly to the preferences of users they
might see samesimilar recommendationsbull Use the right context users do things which might not be relevant for their user model
eg try out things do stuff for other peoplebull Research challenge right balance between serendipity amp personalizationbull Research challenge right way to use the influence of recommendations on userrsquos behavior
Social Web 2016 Lora Aroyo and Davide Ceolin
Announcementsbull Next deadlinesbull Tuesday March 1st 2359
Assignment 2bull Friday March 4th 1000 Post
questionbull Friday March 4th 1700 Vote
questions
image source httpwwwflickrcomphotosbionicteaching1375254387
Hands-on Teaser
bull Your Facebook Friendsrsquo popularity in a spread sheet
bull Locations of your Facebook Friendsbull Tag Cloud of your wall posts
- Social Web 2016
- Slide 2
- Slide 3
- Slide 4
- Slide 5
- Slide 6
- Slide 7
- Slide 8
- Slide 9
- Slide 10
- Slide 11
- Slide 12
- Slide 13
- Slide 14
- Slide 15
- Slide 16
- Slide 17
- Slide 18
- Slide 19
- Slide 20
- Slide 21
- Slide 22
- Slide 23
- Slide 24
- Slide 25
- Slide 26
- Slide 27
- Slide 28
- Slide 29
- Slide 30
- Slide 31
- Slide 32
- Slide 33
- Slide 34
- Slide 35
- Slide 36
- Slide 37
- Slide 38
- Slide 39
- Slide 40
- Slide 41
- Slide 42
- Slide 43
- Slide 44
- Slide 45
- Slide 46
- Slide 47
- Slide 48
- Announcements
- Slide 50
-
Observationsbull Profile characteristics
bull Semantic enrichment solves sparsity problemsbull Profiles change over time recent profiles reflect
better current user demandsbull Temporal patterns weekend profiles differ
significantly from weekday profilesbull Impact on recommendations
bull The more fine-grained the concepts the better the recommendation performance entity-based gt topic-based gt hashtag-based
bull Semantic enrichment improves recommendation quality
bull Time-sensitivity (adapting to trends) improves performance
Social Web 2016 Lora Aroyo and Davide Ceolin
User Modelingit is not about putting everything in a user profile
it is about making the right choices
Social Web 2016 Lora Aroyo and Davide Ceolin
User Adaptation
Knowing the user to adapt a system or interfaceto improve the system functionality and user
experience
Social Web 2016 Lora Aroyo and Davide Ceolin
A Jameson Adaptive interfaces and agents The HCI handbook fundamentals evolving technologies and emerging applications pp 305ndash330 2003
User-Adaptive Systems
Social Web 2016 Lora Aroyo and Davide Ceolin
based on slides from Fabien Abel
Lastfm adapts to your music taste
bull Overfitting ldquobubble effectsrdquo loss of serendipity problem bull systems may adapt too strongly to the
interestsbehaviorbull eg an adaptive radio station may always play the
same or very similar songsbull We search for the right balance between novelty and
relevance for the userbull ldquoLost in Hyperspacerdquo problem
bull when adapting the navigation ndash ie the links on which users can click to findaccess information
bull eg re-orderinghiding of menu items may lead to confusion
Issues in User-Adaptive Systems
Social Web 2016 Lora Aroyo and Davide Ceolin
What is good user modelling amp
personalisation
httpwwwflickrcomphotosbellarosebyliz4729613108
From the consumer perspective of an adaptive system
From the provider perspective of an adaptive system
Success Perspectives
Social Web 2016 Lora Aroyo and Davide Ceolin
bull User studies askobserve (selected) people whether you did a good job
bull Log analysis Analyze (click) data and infer whether you did a good job
bull Evaluation of user modelingbull measure quality of profiles directly eg measure
overlap with existing (true) profiles or let people judge the quality of the generated user profiles
bull measure quality of application that exploits the user profile eg apply user modeling strategies in a recommender system
Evaluation Strategies
Social Web 2016 Lora Aroyo and Davide Ceolin
Evaluating User Modeling in RecSys
Social Web 2016 Lora Aroyo and Davide Ceolin
Possible Metrics
Social Web 2016 Lora Aroyo and Davide Ceolin
bull The usual IR metricsbull Precision fraction of retrieved items that are relevantbull Recall fraction of relevant items that have been
retrievedbull F-Measure (harmonic) mean of precision and recall
bull Metrics for evaluating recommendation (rankings)bull Mean Reciprocal Rank (MRR) of first relevant itembull Successk probability that relevant item occurs
within the top kbull If a true ranking is given rank correlations bull Precisionk Recallk amp F-Measurek
bull Metrics for evaluating prediction of user preferencesbull MAE = Mean Absolute Errorbull TrueFalse PositivesNegatives
Predict relevantusefulinteresting items for a given user (in a given context) itrsquos often a ranking task
Recommendation Systems
Social Web 2016 Lora Aroyo and Davide Ceolin
March 28 2013
Social Web 2016 Lora Aroyo and Davide Ceolin
Social Web 2016 Lora Aroyo and Davide Ceolin
Social Web 2016 Lora Aroyo and Davide Ceolin
httpwwwwiredcommagazine201111mf_artsyall1
Social Web 2016 Lora Aroyo and Davide Ceolin
Collaborative Filteringbull Memory-based User-Item matrix ratingspreferences of users =gt
compute similarity between users amp recommend items of similar users
bull Model-based Item-Item matrix similarity (eg based on user ratings) between items =gt recommend items that are similar to the ones the user likes
bull Model-based Clustering cluster users according to their preferences =gt recommend items of users that belong to the same cluster
bull Model-based Bayesian networks P(u likes item B | u likes item A) = how likely is it that a user who likes item A will like item B learn probabilities from user ratingspreferences
bull Others rule-based other data mining techniques
Social Web 2016 Lora Aroyo and Davide Ceolin
bull complete input data is required
bull pre-computation not possible
bull does not scale well bull high quality of
recommendations
bull abstraction (model) of input data
bull pre-computation (partially) possible (model has to be re-built from time to time)
bull scales betterbull abstraction may reduce
recommendation quality
Memory vs Model-based
Social Web 2016 Lora Aroyo and Davide Ceolin
bull collaborative filtering lsquoneighborhoodsrsquo of people with similar interest amp recommending items based on likings in neighborhood
bull limitations next to lsquocold startrsquo and lsquosparsityrsquo the lack of control (over onersquos neighborhood) is also a problem ie cannot add lsquotrustedrsquo people nor exclude lsquostrangersquo ones
bull therefore interest in lsquosocial recommendersrsquo where presence of social connections defines the similarity in interests (eg social tagging CiteULike)bull does a social connection indicate user interest similaritybull how much users interest similarity depends on the strength
of their connectionbull is it feasible to use a social network as a personalized
recommendation[Lin amp Brusilovsky Social Networks and Interest Similarity The Case of CiteULike HTrsquo10]
Social Networks amp Interest Similarity
bull unilaterally connected pairs have more common itemsmetadatatags than non-connected pairsbull highest similarity for direct connections - decreasing with the increase of distance between users in SN bull reciprocal relationship users - significantly larger similarity than users in a unidirectional relationship bull traditional item-level similarity may be less reliable to find similar users in social bookmarking systemsbull peers connected by self-defined social connections could be a useful source for cross-recommendation
Conclusions
Social Web 2016 Lora Aroyo and Davide Ceolin
bull Input characteristics of items amp interests of a user into characteristics of items =gt Recommend items that feature characteristics which meet the userrsquos interests
bull Techniquesbull Data mining methods Cluster items based on their
characteristics =gt Infer usersrsquo interests into clustersbull IR methods Represent items amp users as term vectors
=gt Compute similarity between user profile vector and items
bull Utility-based methods Utility function that gets an item as input the parameters of the utility function are customized via preferences of a user
Content-based Recommendations
Social Web 2016 Lora Aroyo and Davide Ceolin
based on slides from Fabien Abel
Content Features
based on slides from Fabien Abel
User Model
based on slides from Fabien Abel
Recommendations
RecSys Issuesbull Cold-start problem (new user problem) nolittle data available to infer preferences of new
usersbull Changing User Preferences user interests may change over timebull Sparsity problem (new item problem) item descriptions are sparse eg not many user
rated or tagged an itembull Lack of Diversity (overfitting) when adapting too strongly to the preferences of users they
might see samesimilar recommendationsbull Use the right context users do things which might not be relevant for their user model
eg try out things do stuff for other peoplebull Research challenge right balance between serendipity amp personalizationbull Research challenge right way to use the influence of recommendations on userrsquos behavior
Social Web 2016 Lora Aroyo and Davide Ceolin
Announcementsbull Next deadlinesbull Tuesday March 1st 2359
Assignment 2bull Friday March 4th 1000 Post
questionbull Friday March 4th 1700 Vote
questions
image source httpwwwflickrcomphotosbionicteaching1375254387
Hands-on Teaser
bull Your Facebook Friendsrsquo popularity in a spread sheet
bull Locations of your Facebook Friendsbull Tag Cloud of your wall posts
- Social Web 2016
- Slide 2
- Slide 3
- Slide 4
- Slide 5
- Slide 6
- Slide 7
- Slide 8
- Slide 9
- Slide 10
- Slide 11
- Slide 12
- Slide 13
- Slide 14
- Slide 15
- Slide 16
- Slide 17
- Slide 18
- Slide 19
- Slide 20
- Slide 21
- Slide 22
- Slide 23
- Slide 24
- Slide 25
- Slide 26
- Slide 27
- Slide 28
- Slide 29
- Slide 30
- Slide 31
- Slide 32
- Slide 33
- Slide 34
- Slide 35
- Slide 36
- Slide 37
- Slide 38
- Slide 39
- Slide 40
- Slide 41
- Slide 42
- Slide 43
- Slide 44
- Slide 45
- Slide 46
- Slide 47
- Slide 48
- Announcements
- Slide 50
-
User Modelingit is not about putting everything in a user profile
it is about making the right choices
Social Web 2016 Lora Aroyo and Davide Ceolin
User Adaptation
Knowing the user to adapt a system or interfaceto improve the system functionality and user
experience
Social Web 2016 Lora Aroyo and Davide Ceolin
A Jameson Adaptive interfaces and agents The HCI handbook fundamentals evolving technologies and emerging applications pp 305ndash330 2003
User-Adaptive Systems
Social Web 2016 Lora Aroyo and Davide Ceolin
based on slides from Fabien Abel
Lastfm adapts to your music taste
bull Overfitting ldquobubble effectsrdquo loss of serendipity problem bull systems may adapt too strongly to the
interestsbehaviorbull eg an adaptive radio station may always play the
same or very similar songsbull We search for the right balance between novelty and
relevance for the userbull ldquoLost in Hyperspacerdquo problem
bull when adapting the navigation ndash ie the links on which users can click to findaccess information
bull eg re-orderinghiding of menu items may lead to confusion
Issues in User-Adaptive Systems
Social Web 2016 Lora Aroyo and Davide Ceolin
What is good user modelling amp
personalisation
httpwwwflickrcomphotosbellarosebyliz4729613108
From the consumer perspective of an adaptive system
From the provider perspective of an adaptive system
Success Perspectives
Social Web 2016 Lora Aroyo and Davide Ceolin
bull User studies askobserve (selected) people whether you did a good job
bull Log analysis Analyze (click) data and infer whether you did a good job
bull Evaluation of user modelingbull measure quality of profiles directly eg measure
overlap with existing (true) profiles or let people judge the quality of the generated user profiles
bull measure quality of application that exploits the user profile eg apply user modeling strategies in a recommender system
Evaluation Strategies
Social Web 2016 Lora Aroyo and Davide Ceolin
Evaluating User Modeling in RecSys
Social Web 2016 Lora Aroyo and Davide Ceolin
Possible Metrics
Social Web 2016 Lora Aroyo and Davide Ceolin
bull The usual IR metricsbull Precision fraction of retrieved items that are relevantbull Recall fraction of relevant items that have been
retrievedbull F-Measure (harmonic) mean of precision and recall
bull Metrics for evaluating recommendation (rankings)bull Mean Reciprocal Rank (MRR) of first relevant itembull Successk probability that relevant item occurs
within the top kbull If a true ranking is given rank correlations bull Precisionk Recallk amp F-Measurek
bull Metrics for evaluating prediction of user preferencesbull MAE = Mean Absolute Errorbull TrueFalse PositivesNegatives
Predict relevantusefulinteresting items for a given user (in a given context) itrsquos often a ranking task
Recommendation Systems
Social Web 2016 Lora Aroyo and Davide Ceolin
March 28 2013
Social Web 2016 Lora Aroyo and Davide Ceolin
Social Web 2016 Lora Aroyo and Davide Ceolin
Social Web 2016 Lora Aroyo and Davide Ceolin
httpwwwwiredcommagazine201111mf_artsyall1
Social Web 2016 Lora Aroyo and Davide Ceolin
Collaborative Filteringbull Memory-based User-Item matrix ratingspreferences of users =gt
compute similarity between users amp recommend items of similar users
bull Model-based Item-Item matrix similarity (eg based on user ratings) between items =gt recommend items that are similar to the ones the user likes
bull Model-based Clustering cluster users according to their preferences =gt recommend items of users that belong to the same cluster
bull Model-based Bayesian networks P(u likes item B | u likes item A) = how likely is it that a user who likes item A will like item B learn probabilities from user ratingspreferences
bull Others rule-based other data mining techniques
Social Web 2016 Lora Aroyo and Davide Ceolin
bull complete input data is required
bull pre-computation not possible
bull does not scale well bull high quality of
recommendations
bull abstraction (model) of input data
bull pre-computation (partially) possible (model has to be re-built from time to time)
bull scales betterbull abstraction may reduce
recommendation quality
Memory vs Model-based
Social Web 2016 Lora Aroyo and Davide Ceolin
bull collaborative filtering lsquoneighborhoodsrsquo of people with similar interest amp recommending items based on likings in neighborhood
bull limitations next to lsquocold startrsquo and lsquosparsityrsquo the lack of control (over onersquos neighborhood) is also a problem ie cannot add lsquotrustedrsquo people nor exclude lsquostrangersquo ones
bull therefore interest in lsquosocial recommendersrsquo where presence of social connections defines the similarity in interests (eg social tagging CiteULike)bull does a social connection indicate user interest similaritybull how much users interest similarity depends on the strength
of their connectionbull is it feasible to use a social network as a personalized
recommendation[Lin amp Brusilovsky Social Networks and Interest Similarity The Case of CiteULike HTrsquo10]
Social Networks amp Interest Similarity
bull unilaterally connected pairs have more common itemsmetadatatags than non-connected pairsbull highest similarity for direct connections - decreasing with the increase of distance between users in SN bull reciprocal relationship users - significantly larger similarity than users in a unidirectional relationship bull traditional item-level similarity may be less reliable to find similar users in social bookmarking systemsbull peers connected by self-defined social connections could be a useful source for cross-recommendation
Conclusions
Social Web 2016 Lora Aroyo and Davide Ceolin
bull Input characteristics of items amp interests of a user into characteristics of items =gt Recommend items that feature characteristics which meet the userrsquos interests
bull Techniquesbull Data mining methods Cluster items based on their
characteristics =gt Infer usersrsquo interests into clustersbull IR methods Represent items amp users as term vectors
=gt Compute similarity between user profile vector and items
bull Utility-based methods Utility function that gets an item as input the parameters of the utility function are customized via preferences of a user
Content-based Recommendations
Social Web 2016 Lora Aroyo and Davide Ceolin
based on slides from Fabien Abel
Content Features
based on slides from Fabien Abel
User Model
based on slides from Fabien Abel
Recommendations
RecSys Issuesbull Cold-start problem (new user problem) nolittle data available to infer preferences of new
usersbull Changing User Preferences user interests may change over timebull Sparsity problem (new item problem) item descriptions are sparse eg not many user
rated or tagged an itembull Lack of Diversity (overfitting) when adapting too strongly to the preferences of users they
might see samesimilar recommendationsbull Use the right context users do things which might not be relevant for their user model
eg try out things do stuff for other peoplebull Research challenge right balance between serendipity amp personalizationbull Research challenge right way to use the influence of recommendations on userrsquos behavior
Social Web 2016 Lora Aroyo and Davide Ceolin
Announcementsbull Next deadlinesbull Tuesday March 1st 2359
Assignment 2bull Friday March 4th 1000 Post
questionbull Friday March 4th 1700 Vote
questions
image source httpwwwflickrcomphotosbionicteaching1375254387
Hands-on Teaser
bull Your Facebook Friendsrsquo popularity in a spread sheet
bull Locations of your Facebook Friendsbull Tag Cloud of your wall posts
- Social Web 2016
- Slide 2
- Slide 3
- Slide 4
- Slide 5
- Slide 6
- Slide 7
- Slide 8
- Slide 9
- Slide 10
- Slide 11
- Slide 12
- Slide 13
- Slide 14
- Slide 15
- Slide 16
- Slide 17
- Slide 18
- Slide 19
- Slide 20
- Slide 21
- Slide 22
- Slide 23
- Slide 24
- Slide 25
- Slide 26
- Slide 27
- Slide 28
- Slide 29
- Slide 30
- Slide 31
- Slide 32
- Slide 33
- Slide 34
- Slide 35
- Slide 36
- Slide 37
- Slide 38
- Slide 39
- Slide 40
- Slide 41
- Slide 42
- Slide 43
- Slide 44
- Slide 45
- Slide 46
- Slide 47
- Slide 48
- Announcements
- Slide 50
-
User Adaptation
Knowing the user to adapt a system or interfaceto improve the system functionality and user
experience
Social Web 2016 Lora Aroyo and Davide Ceolin
A Jameson Adaptive interfaces and agents The HCI handbook fundamentals evolving technologies and emerging applications pp 305ndash330 2003
User-Adaptive Systems
Social Web 2016 Lora Aroyo and Davide Ceolin
based on slides from Fabien Abel
Lastfm adapts to your music taste
bull Overfitting ldquobubble effectsrdquo loss of serendipity problem bull systems may adapt too strongly to the
interestsbehaviorbull eg an adaptive radio station may always play the
same or very similar songsbull We search for the right balance between novelty and
relevance for the userbull ldquoLost in Hyperspacerdquo problem
bull when adapting the navigation ndash ie the links on which users can click to findaccess information
bull eg re-orderinghiding of menu items may lead to confusion
Issues in User-Adaptive Systems
Social Web 2016 Lora Aroyo and Davide Ceolin
What is good user modelling amp
personalisation
httpwwwflickrcomphotosbellarosebyliz4729613108
From the consumer perspective of an adaptive system
From the provider perspective of an adaptive system
Success Perspectives
Social Web 2016 Lora Aroyo and Davide Ceolin
bull User studies askobserve (selected) people whether you did a good job
bull Log analysis Analyze (click) data and infer whether you did a good job
bull Evaluation of user modelingbull measure quality of profiles directly eg measure
overlap with existing (true) profiles or let people judge the quality of the generated user profiles
bull measure quality of application that exploits the user profile eg apply user modeling strategies in a recommender system
Evaluation Strategies
Social Web 2016 Lora Aroyo and Davide Ceolin
Evaluating User Modeling in RecSys
Social Web 2016 Lora Aroyo and Davide Ceolin
Possible Metrics
Social Web 2016 Lora Aroyo and Davide Ceolin
bull The usual IR metricsbull Precision fraction of retrieved items that are relevantbull Recall fraction of relevant items that have been
retrievedbull F-Measure (harmonic) mean of precision and recall
bull Metrics for evaluating recommendation (rankings)bull Mean Reciprocal Rank (MRR) of first relevant itembull Successk probability that relevant item occurs
within the top kbull If a true ranking is given rank correlations bull Precisionk Recallk amp F-Measurek
bull Metrics for evaluating prediction of user preferencesbull MAE = Mean Absolute Errorbull TrueFalse PositivesNegatives
Predict relevantusefulinteresting items for a given user (in a given context) itrsquos often a ranking task
Recommendation Systems
Social Web 2016 Lora Aroyo and Davide Ceolin
March 28 2013
Social Web 2016 Lora Aroyo and Davide Ceolin
Social Web 2016 Lora Aroyo and Davide Ceolin
Social Web 2016 Lora Aroyo and Davide Ceolin
httpwwwwiredcommagazine201111mf_artsyall1
Social Web 2016 Lora Aroyo and Davide Ceolin
Collaborative Filteringbull Memory-based User-Item matrix ratingspreferences of users =gt
compute similarity between users amp recommend items of similar users
bull Model-based Item-Item matrix similarity (eg based on user ratings) between items =gt recommend items that are similar to the ones the user likes
bull Model-based Clustering cluster users according to their preferences =gt recommend items of users that belong to the same cluster
bull Model-based Bayesian networks P(u likes item B | u likes item A) = how likely is it that a user who likes item A will like item B learn probabilities from user ratingspreferences
bull Others rule-based other data mining techniques
Social Web 2016 Lora Aroyo and Davide Ceolin
bull complete input data is required
bull pre-computation not possible
bull does not scale well bull high quality of
recommendations
bull abstraction (model) of input data
bull pre-computation (partially) possible (model has to be re-built from time to time)
bull scales betterbull abstraction may reduce
recommendation quality
Memory vs Model-based
Social Web 2016 Lora Aroyo and Davide Ceolin
bull collaborative filtering lsquoneighborhoodsrsquo of people with similar interest amp recommending items based on likings in neighborhood
bull limitations next to lsquocold startrsquo and lsquosparsityrsquo the lack of control (over onersquos neighborhood) is also a problem ie cannot add lsquotrustedrsquo people nor exclude lsquostrangersquo ones
bull therefore interest in lsquosocial recommendersrsquo where presence of social connections defines the similarity in interests (eg social tagging CiteULike)bull does a social connection indicate user interest similaritybull how much users interest similarity depends on the strength
of their connectionbull is it feasible to use a social network as a personalized
recommendation[Lin amp Brusilovsky Social Networks and Interest Similarity The Case of CiteULike HTrsquo10]
Social Networks amp Interest Similarity
bull unilaterally connected pairs have more common itemsmetadatatags than non-connected pairsbull highest similarity for direct connections - decreasing with the increase of distance between users in SN bull reciprocal relationship users - significantly larger similarity than users in a unidirectional relationship bull traditional item-level similarity may be less reliable to find similar users in social bookmarking systemsbull peers connected by self-defined social connections could be a useful source for cross-recommendation
Conclusions
Social Web 2016 Lora Aroyo and Davide Ceolin
bull Input characteristics of items amp interests of a user into characteristics of items =gt Recommend items that feature characteristics which meet the userrsquos interests
bull Techniquesbull Data mining methods Cluster items based on their
characteristics =gt Infer usersrsquo interests into clustersbull IR methods Represent items amp users as term vectors
=gt Compute similarity between user profile vector and items
bull Utility-based methods Utility function that gets an item as input the parameters of the utility function are customized via preferences of a user
Content-based Recommendations
Social Web 2016 Lora Aroyo and Davide Ceolin
based on slides from Fabien Abel
Content Features
based on slides from Fabien Abel
User Model
based on slides from Fabien Abel
Recommendations
RecSys Issuesbull Cold-start problem (new user problem) nolittle data available to infer preferences of new
usersbull Changing User Preferences user interests may change over timebull Sparsity problem (new item problem) item descriptions are sparse eg not many user
rated or tagged an itembull Lack of Diversity (overfitting) when adapting too strongly to the preferences of users they
might see samesimilar recommendationsbull Use the right context users do things which might not be relevant for their user model
eg try out things do stuff for other peoplebull Research challenge right balance between serendipity amp personalizationbull Research challenge right way to use the influence of recommendations on userrsquos behavior
Social Web 2016 Lora Aroyo and Davide Ceolin
Announcementsbull Next deadlinesbull Tuesday March 1st 2359
Assignment 2bull Friday March 4th 1000 Post
questionbull Friday March 4th 1700 Vote
questions
image source httpwwwflickrcomphotosbionicteaching1375254387
Hands-on Teaser
bull Your Facebook Friendsrsquo popularity in a spread sheet
bull Locations of your Facebook Friendsbull Tag Cloud of your wall posts
- Social Web 2016
- Slide 2
- Slide 3
- Slide 4
- Slide 5
- Slide 6
- Slide 7
- Slide 8
- Slide 9
- Slide 10
- Slide 11
- Slide 12
- Slide 13
- Slide 14
- Slide 15
- Slide 16
- Slide 17
- Slide 18
- Slide 19
- Slide 20
- Slide 21
- Slide 22
- Slide 23
- Slide 24
- Slide 25
- Slide 26
- Slide 27
- Slide 28
- Slide 29
- Slide 30
- Slide 31
- Slide 32
- Slide 33
- Slide 34
- Slide 35
- Slide 36
- Slide 37
- Slide 38
- Slide 39
- Slide 40
- Slide 41
- Slide 42
- Slide 43
- Slide 44
- Slide 45
- Slide 46
- Slide 47
- Slide 48
- Announcements
- Slide 50
-
A Jameson Adaptive interfaces and agents The HCI handbook fundamentals evolving technologies and emerging applications pp 305ndash330 2003
User-Adaptive Systems
Social Web 2016 Lora Aroyo and Davide Ceolin
based on slides from Fabien Abel
Lastfm adapts to your music taste
bull Overfitting ldquobubble effectsrdquo loss of serendipity problem bull systems may adapt too strongly to the
interestsbehaviorbull eg an adaptive radio station may always play the
same or very similar songsbull We search for the right balance between novelty and
relevance for the userbull ldquoLost in Hyperspacerdquo problem
bull when adapting the navigation ndash ie the links on which users can click to findaccess information
bull eg re-orderinghiding of menu items may lead to confusion
Issues in User-Adaptive Systems
Social Web 2016 Lora Aroyo and Davide Ceolin
What is good user modelling amp
personalisation
httpwwwflickrcomphotosbellarosebyliz4729613108
From the consumer perspective of an adaptive system
From the provider perspective of an adaptive system
Success Perspectives
Social Web 2016 Lora Aroyo and Davide Ceolin
bull User studies askobserve (selected) people whether you did a good job
bull Log analysis Analyze (click) data and infer whether you did a good job
bull Evaluation of user modelingbull measure quality of profiles directly eg measure
overlap with existing (true) profiles or let people judge the quality of the generated user profiles
bull measure quality of application that exploits the user profile eg apply user modeling strategies in a recommender system
Evaluation Strategies
Social Web 2016 Lora Aroyo and Davide Ceolin
Evaluating User Modeling in RecSys
Social Web 2016 Lora Aroyo and Davide Ceolin
Possible Metrics
Social Web 2016 Lora Aroyo and Davide Ceolin
bull The usual IR metricsbull Precision fraction of retrieved items that are relevantbull Recall fraction of relevant items that have been
retrievedbull F-Measure (harmonic) mean of precision and recall
bull Metrics for evaluating recommendation (rankings)bull Mean Reciprocal Rank (MRR) of first relevant itembull Successk probability that relevant item occurs
within the top kbull If a true ranking is given rank correlations bull Precisionk Recallk amp F-Measurek
bull Metrics for evaluating prediction of user preferencesbull MAE = Mean Absolute Errorbull TrueFalse PositivesNegatives
Predict relevantusefulinteresting items for a given user (in a given context) itrsquos often a ranking task
Recommendation Systems
Social Web 2016 Lora Aroyo and Davide Ceolin
March 28 2013
Social Web 2016 Lora Aroyo and Davide Ceolin
Social Web 2016 Lora Aroyo and Davide Ceolin
Social Web 2016 Lora Aroyo and Davide Ceolin
httpwwwwiredcommagazine201111mf_artsyall1
Social Web 2016 Lora Aroyo and Davide Ceolin
Collaborative Filteringbull Memory-based User-Item matrix ratingspreferences of users =gt
compute similarity between users amp recommend items of similar users
bull Model-based Item-Item matrix similarity (eg based on user ratings) between items =gt recommend items that are similar to the ones the user likes
bull Model-based Clustering cluster users according to their preferences =gt recommend items of users that belong to the same cluster
bull Model-based Bayesian networks P(u likes item B | u likes item A) = how likely is it that a user who likes item A will like item B learn probabilities from user ratingspreferences
bull Others rule-based other data mining techniques
Social Web 2016 Lora Aroyo and Davide Ceolin
bull complete input data is required
bull pre-computation not possible
bull does not scale well bull high quality of
recommendations
bull abstraction (model) of input data
bull pre-computation (partially) possible (model has to be re-built from time to time)
bull scales betterbull abstraction may reduce
recommendation quality
Memory vs Model-based
Social Web 2016 Lora Aroyo and Davide Ceolin
bull collaborative filtering lsquoneighborhoodsrsquo of people with similar interest amp recommending items based on likings in neighborhood
bull limitations next to lsquocold startrsquo and lsquosparsityrsquo the lack of control (over onersquos neighborhood) is also a problem ie cannot add lsquotrustedrsquo people nor exclude lsquostrangersquo ones
bull therefore interest in lsquosocial recommendersrsquo where presence of social connections defines the similarity in interests (eg social tagging CiteULike)bull does a social connection indicate user interest similaritybull how much users interest similarity depends on the strength
of their connectionbull is it feasible to use a social network as a personalized
recommendation[Lin amp Brusilovsky Social Networks and Interest Similarity The Case of CiteULike HTrsquo10]
Social Networks amp Interest Similarity
bull unilaterally connected pairs have more common itemsmetadatatags than non-connected pairsbull highest similarity for direct connections - decreasing with the increase of distance between users in SN bull reciprocal relationship users - significantly larger similarity than users in a unidirectional relationship bull traditional item-level similarity may be less reliable to find similar users in social bookmarking systemsbull peers connected by self-defined social connections could be a useful source for cross-recommendation
Conclusions
Social Web 2016 Lora Aroyo and Davide Ceolin
bull Input characteristics of items amp interests of a user into characteristics of items =gt Recommend items that feature characteristics which meet the userrsquos interests
bull Techniquesbull Data mining methods Cluster items based on their
characteristics =gt Infer usersrsquo interests into clustersbull IR methods Represent items amp users as term vectors
=gt Compute similarity between user profile vector and items
bull Utility-based methods Utility function that gets an item as input the parameters of the utility function are customized via preferences of a user
Content-based Recommendations
Social Web 2016 Lora Aroyo and Davide Ceolin
based on slides from Fabien Abel
Content Features
based on slides from Fabien Abel
User Model
based on slides from Fabien Abel
Recommendations
RecSys Issuesbull Cold-start problem (new user problem) nolittle data available to infer preferences of new
usersbull Changing User Preferences user interests may change over timebull Sparsity problem (new item problem) item descriptions are sparse eg not many user
rated or tagged an itembull Lack of Diversity (overfitting) when adapting too strongly to the preferences of users they
might see samesimilar recommendationsbull Use the right context users do things which might not be relevant for their user model
eg try out things do stuff for other peoplebull Research challenge right balance between serendipity amp personalizationbull Research challenge right way to use the influence of recommendations on userrsquos behavior
Social Web 2016 Lora Aroyo and Davide Ceolin
Announcementsbull Next deadlinesbull Tuesday March 1st 2359
Assignment 2bull Friday March 4th 1000 Post
questionbull Friday March 4th 1700 Vote
questions
image source httpwwwflickrcomphotosbionicteaching1375254387
Hands-on Teaser
bull Your Facebook Friendsrsquo popularity in a spread sheet
bull Locations of your Facebook Friendsbull Tag Cloud of your wall posts
- Social Web 2016
- Slide 2
- Slide 3
- Slide 4
- Slide 5
- Slide 6
- Slide 7
- Slide 8
- Slide 9
- Slide 10
- Slide 11
- Slide 12
- Slide 13
- Slide 14
- Slide 15
- Slide 16
- Slide 17
- Slide 18
- Slide 19
- Slide 20
- Slide 21
- Slide 22
- Slide 23
- Slide 24
- Slide 25
- Slide 26
- Slide 27
- Slide 28
- Slide 29
- Slide 30
- Slide 31
- Slide 32
- Slide 33
- Slide 34
- Slide 35
- Slide 36
- Slide 37
- Slide 38
- Slide 39
- Slide 40
- Slide 41
- Slide 42
- Slide 43
- Slide 44
- Slide 45
- Slide 46
- Slide 47
- Slide 48
- Announcements
- Slide 50
-
based on slides from Fabien Abel
Lastfm adapts to your music taste
bull Overfitting ldquobubble effectsrdquo loss of serendipity problem bull systems may adapt too strongly to the
interestsbehaviorbull eg an adaptive radio station may always play the
same or very similar songsbull We search for the right balance between novelty and
relevance for the userbull ldquoLost in Hyperspacerdquo problem
bull when adapting the navigation ndash ie the links on which users can click to findaccess information
bull eg re-orderinghiding of menu items may lead to confusion
Issues in User-Adaptive Systems
Social Web 2016 Lora Aroyo and Davide Ceolin
What is good user modelling amp
personalisation
httpwwwflickrcomphotosbellarosebyliz4729613108
From the consumer perspective of an adaptive system
From the provider perspective of an adaptive system
Success Perspectives
Social Web 2016 Lora Aroyo and Davide Ceolin
bull User studies askobserve (selected) people whether you did a good job
bull Log analysis Analyze (click) data and infer whether you did a good job
bull Evaluation of user modelingbull measure quality of profiles directly eg measure
overlap with existing (true) profiles or let people judge the quality of the generated user profiles
bull measure quality of application that exploits the user profile eg apply user modeling strategies in a recommender system
Evaluation Strategies
Social Web 2016 Lora Aroyo and Davide Ceolin
Evaluating User Modeling in RecSys
Social Web 2016 Lora Aroyo and Davide Ceolin
Possible Metrics
Social Web 2016 Lora Aroyo and Davide Ceolin
bull The usual IR metricsbull Precision fraction of retrieved items that are relevantbull Recall fraction of relevant items that have been
retrievedbull F-Measure (harmonic) mean of precision and recall
bull Metrics for evaluating recommendation (rankings)bull Mean Reciprocal Rank (MRR) of first relevant itembull Successk probability that relevant item occurs
within the top kbull If a true ranking is given rank correlations bull Precisionk Recallk amp F-Measurek
bull Metrics for evaluating prediction of user preferencesbull MAE = Mean Absolute Errorbull TrueFalse PositivesNegatives
Predict relevantusefulinteresting items for a given user (in a given context) itrsquos often a ranking task
Recommendation Systems
Social Web 2016 Lora Aroyo and Davide Ceolin
March 28 2013
Social Web 2016 Lora Aroyo and Davide Ceolin
Social Web 2016 Lora Aroyo and Davide Ceolin
Social Web 2016 Lora Aroyo and Davide Ceolin
httpwwwwiredcommagazine201111mf_artsyall1
Social Web 2016 Lora Aroyo and Davide Ceolin
Collaborative Filteringbull Memory-based User-Item matrix ratingspreferences of users =gt
compute similarity between users amp recommend items of similar users
bull Model-based Item-Item matrix similarity (eg based on user ratings) between items =gt recommend items that are similar to the ones the user likes
bull Model-based Clustering cluster users according to their preferences =gt recommend items of users that belong to the same cluster
bull Model-based Bayesian networks P(u likes item B | u likes item A) = how likely is it that a user who likes item A will like item B learn probabilities from user ratingspreferences
bull Others rule-based other data mining techniques
Social Web 2016 Lora Aroyo and Davide Ceolin
bull complete input data is required
bull pre-computation not possible
bull does not scale well bull high quality of
recommendations
bull abstraction (model) of input data
bull pre-computation (partially) possible (model has to be re-built from time to time)
bull scales betterbull abstraction may reduce
recommendation quality
Memory vs Model-based
Social Web 2016 Lora Aroyo and Davide Ceolin
bull collaborative filtering lsquoneighborhoodsrsquo of people with similar interest amp recommending items based on likings in neighborhood
bull limitations next to lsquocold startrsquo and lsquosparsityrsquo the lack of control (over onersquos neighborhood) is also a problem ie cannot add lsquotrustedrsquo people nor exclude lsquostrangersquo ones
bull therefore interest in lsquosocial recommendersrsquo where presence of social connections defines the similarity in interests (eg social tagging CiteULike)bull does a social connection indicate user interest similaritybull how much users interest similarity depends on the strength
of their connectionbull is it feasible to use a social network as a personalized
recommendation[Lin amp Brusilovsky Social Networks and Interest Similarity The Case of CiteULike HTrsquo10]
Social Networks amp Interest Similarity
bull unilaterally connected pairs have more common itemsmetadatatags than non-connected pairsbull highest similarity for direct connections - decreasing with the increase of distance between users in SN bull reciprocal relationship users - significantly larger similarity than users in a unidirectional relationship bull traditional item-level similarity may be less reliable to find similar users in social bookmarking systemsbull peers connected by self-defined social connections could be a useful source for cross-recommendation
Conclusions
Social Web 2016 Lora Aroyo and Davide Ceolin
bull Input characteristics of items amp interests of a user into characteristics of items =gt Recommend items that feature characteristics which meet the userrsquos interests
bull Techniquesbull Data mining methods Cluster items based on their
characteristics =gt Infer usersrsquo interests into clustersbull IR methods Represent items amp users as term vectors
=gt Compute similarity between user profile vector and items
bull Utility-based methods Utility function that gets an item as input the parameters of the utility function are customized via preferences of a user
Content-based Recommendations
Social Web 2016 Lora Aroyo and Davide Ceolin
based on slides from Fabien Abel
Content Features
based on slides from Fabien Abel
User Model
based on slides from Fabien Abel
Recommendations
RecSys Issuesbull Cold-start problem (new user problem) nolittle data available to infer preferences of new
usersbull Changing User Preferences user interests may change over timebull Sparsity problem (new item problem) item descriptions are sparse eg not many user
rated or tagged an itembull Lack of Diversity (overfitting) when adapting too strongly to the preferences of users they
might see samesimilar recommendationsbull Use the right context users do things which might not be relevant for their user model
eg try out things do stuff for other peoplebull Research challenge right balance between serendipity amp personalizationbull Research challenge right way to use the influence of recommendations on userrsquos behavior
Social Web 2016 Lora Aroyo and Davide Ceolin
Announcementsbull Next deadlinesbull Tuesday March 1st 2359
Assignment 2bull Friday March 4th 1000 Post
questionbull Friday March 4th 1700 Vote
questions
image source httpwwwflickrcomphotosbionicteaching1375254387
Hands-on Teaser
bull Your Facebook Friendsrsquo popularity in a spread sheet
bull Locations of your Facebook Friendsbull Tag Cloud of your wall posts
- Social Web 2016
- Slide 2
- Slide 3
- Slide 4
- Slide 5
- Slide 6
- Slide 7
- Slide 8
- Slide 9
- Slide 10
- Slide 11
- Slide 12
- Slide 13
- Slide 14
- Slide 15
- Slide 16
- Slide 17
- Slide 18
- Slide 19
- Slide 20
- Slide 21
- Slide 22
- Slide 23
- Slide 24
- Slide 25
- Slide 26
- Slide 27
- Slide 28
- Slide 29
- Slide 30
- Slide 31
- Slide 32
- Slide 33
- Slide 34
- Slide 35
- Slide 36
- Slide 37
- Slide 38
- Slide 39
- Slide 40
- Slide 41
- Slide 42
- Slide 43
- Slide 44
- Slide 45
- Slide 46
- Slide 47
- Slide 48
- Announcements
- Slide 50
-
bull Overfitting ldquobubble effectsrdquo loss of serendipity problem bull systems may adapt too strongly to the
interestsbehaviorbull eg an adaptive radio station may always play the
same or very similar songsbull We search for the right balance between novelty and
relevance for the userbull ldquoLost in Hyperspacerdquo problem
bull when adapting the navigation ndash ie the links on which users can click to findaccess information
bull eg re-orderinghiding of menu items may lead to confusion
Issues in User-Adaptive Systems
Social Web 2016 Lora Aroyo and Davide Ceolin
What is good user modelling amp
personalisation
httpwwwflickrcomphotosbellarosebyliz4729613108
From the consumer perspective of an adaptive system
From the provider perspective of an adaptive system
Success Perspectives
Social Web 2016 Lora Aroyo and Davide Ceolin
bull User studies askobserve (selected) people whether you did a good job
bull Log analysis Analyze (click) data and infer whether you did a good job
bull Evaluation of user modelingbull measure quality of profiles directly eg measure
overlap with existing (true) profiles or let people judge the quality of the generated user profiles
bull measure quality of application that exploits the user profile eg apply user modeling strategies in a recommender system
Evaluation Strategies
Social Web 2016 Lora Aroyo and Davide Ceolin
Evaluating User Modeling in RecSys
Social Web 2016 Lora Aroyo and Davide Ceolin
Possible Metrics
Social Web 2016 Lora Aroyo and Davide Ceolin
bull The usual IR metricsbull Precision fraction of retrieved items that are relevantbull Recall fraction of relevant items that have been
retrievedbull F-Measure (harmonic) mean of precision and recall
bull Metrics for evaluating recommendation (rankings)bull Mean Reciprocal Rank (MRR) of first relevant itembull Successk probability that relevant item occurs
within the top kbull If a true ranking is given rank correlations bull Precisionk Recallk amp F-Measurek
bull Metrics for evaluating prediction of user preferencesbull MAE = Mean Absolute Errorbull TrueFalse PositivesNegatives
Predict relevantusefulinteresting items for a given user (in a given context) itrsquos often a ranking task
Recommendation Systems
Social Web 2016 Lora Aroyo and Davide Ceolin
March 28 2013
Social Web 2016 Lora Aroyo and Davide Ceolin
Social Web 2016 Lora Aroyo and Davide Ceolin
Social Web 2016 Lora Aroyo and Davide Ceolin
httpwwwwiredcommagazine201111mf_artsyall1
Social Web 2016 Lora Aroyo and Davide Ceolin
Collaborative Filteringbull Memory-based User-Item matrix ratingspreferences of users =gt
compute similarity between users amp recommend items of similar users
bull Model-based Item-Item matrix similarity (eg based on user ratings) between items =gt recommend items that are similar to the ones the user likes
bull Model-based Clustering cluster users according to their preferences =gt recommend items of users that belong to the same cluster
bull Model-based Bayesian networks P(u likes item B | u likes item A) = how likely is it that a user who likes item A will like item B learn probabilities from user ratingspreferences
bull Others rule-based other data mining techniques
Social Web 2016 Lora Aroyo and Davide Ceolin
bull complete input data is required
bull pre-computation not possible
bull does not scale well bull high quality of
recommendations
bull abstraction (model) of input data
bull pre-computation (partially) possible (model has to be re-built from time to time)
bull scales betterbull abstraction may reduce
recommendation quality
Memory vs Model-based
Social Web 2016 Lora Aroyo and Davide Ceolin
bull collaborative filtering lsquoneighborhoodsrsquo of people with similar interest amp recommending items based on likings in neighborhood
bull limitations next to lsquocold startrsquo and lsquosparsityrsquo the lack of control (over onersquos neighborhood) is also a problem ie cannot add lsquotrustedrsquo people nor exclude lsquostrangersquo ones
bull therefore interest in lsquosocial recommendersrsquo where presence of social connections defines the similarity in interests (eg social tagging CiteULike)bull does a social connection indicate user interest similaritybull how much users interest similarity depends on the strength
of their connectionbull is it feasible to use a social network as a personalized
recommendation[Lin amp Brusilovsky Social Networks and Interest Similarity The Case of CiteULike HTrsquo10]
Social Networks amp Interest Similarity
bull unilaterally connected pairs have more common itemsmetadatatags than non-connected pairsbull highest similarity for direct connections - decreasing with the increase of distance between users in SN bull reciprocal relationship users - significantly larger similarity than users in a unidirectional relationship bull traditional item-level similarity may be less reliable to find similar users in social bookmarking systemsbull peers connected by self-defined social connections could be a useful source for cross-recommendation
Conclusions
Social Web 2016 Lora Aroyo and Davide Ceolin
bull Input characteristics of items amp interests of a user into characteristics of items =gt Recommend items that feature characteristics which meet the userrsquos interests
bull Techniquesbull Data mining methods Cluster items based on their
characteristics =gt Infer usersrsquo interests into clustersbull IR methods Represent items amp users as term vectors
=gt Compute similarity between user profile vector and items
bull Utility-based methods Utility function that gets an item as input the parameters of the utility function are customized via preferences of a user
Content-based Recommendations
Social Web 2016 Lora Aroyo and Davide Ceolin
based on slides from Fabien Abel
Content Features
based on slides from Fabien Abel
User Model
based on slides from Fabien Abel
Recommendations
RecSys Issuesbull Cold-start problem (new user problem) nolittle data available to infer preferences of new
usersbull Changing User Preferences user interests may change over timebull Sparsity problem (new item problem) item descriptions are sparse eg not many user
rated or tagged an itembull Lack of Diversity (overfitting) when adapting too strongly to the preferences of users they
might see samesimilar recommendationsbull Use the right context users do things which might not be relevant for their user model
eg try out things do stuff for other peoplebull Research challenge right balance between serendipity amp personalizationbull Research challenge right way to use the influence of recommendations on userrsquos behavior
Social Web 2016 Lora Aroyo and Davide Ceolin
Announcementsbull Next deadlinesbull Tuesday March 1st 2359
Assignment 2bull Friday March 4th 1000 Post
questionbull Friday March 4th 1700 Vote
questions
image source httpwwwflickrcomphotosbionicteaching1375254387
Hands-on Teaser
bull Your Facebook Friendsrsquo popularity in a spread sheet
bull Locations of your Facebook Friendsbull Tag Cloud of your wall posts
- Social Web 2016
- Slide 2
- Slide 3
- Slide 4
- Slide 5
- Slide 6
- Slide 7
- Slide 8
- Slide 9
- Slide 10
- Slide 11
- Slide 12
- Slide 13
- Slide 14
- Slide 15
- Slide 16
- Slide 17
- Slide 18
- Slide 19
- Slide 20
- Slide 21
- Slide 22
- Slide 23
- Slide 24
- Slide 25
- Slide 26
- Slide 27
- Slide 28
- Slide 29
- Slide 30
- Slide 31
- Slide 32
- Slide 33
- Slide 34
- Slide 35
- Slide 36
- Slide 37
- Slide 38
- Slide 39
- Slide 40
- Slide 41
- Slide 42
- Slide 43
- Slide 44
- Slide 45
- Slide 46
- Slide 47
- Slide 48
- Announcements
- Slide 50
-
What is good user modelling amp
personalisation
httpwwwflickrcomphotosbellarosebyliz4729613108
From the consumer perspective of an adaptive system
From the provider perspective of an adaptive system
Success Perspectives
Social Web 2016 Lora Aroyo and Davide Ceolin
bull User studies askobserve (selected) people whether you did a good job
bull Log analysis Analyze (click) data and infer whether you did a good job
bull Evaluation of user modelingbull measure quality of profiles directly eg measure
overlap with existing (true) profiles or let people judge the quality of the generated user profiles
bull measure quality of application that exploits the user profile eg apply user modeling strategies in a recommender system
Evaluation Strategies
Social Web 2016 Lora Aroyo and Davide Ceolin
Evaluating User Modeling in RecSys
Social Web 2016 Lora Aroyo and Davide Ceolin
Possible Metrics
Social Web 2016 Lora Aroyo and Davide Ceolin
bull The usual IR metricsbull Precision fraction of retrieved items that are relevantbull Recall fraction of relevant items that have been
retrievedbull F-Measure (harmonic) mean of precision and recall
bull Metrics for evaluating recommendation (rankings)bull Mean Reciprocal Rank (MRR) of first relevant itembull Successk probability that relevant item occurs
within the top kbull If a true ranking is given rank correlations bull Precisionk Recallk amp F-Measurek
bull Metrics for evaluating prediction of user preferencesbull MAE = Mean Absolute Errorbull TrueFalse PositivesNegatives
Predict relevantusefulinteresting items for a given user (in a given context) itrsquos often a ranking task
Recommendation Systems
Social Web 2016 Lora Aroyo and Davide Ceolin
March 28 2013
Social Web 2016 Lora Aroyo and Davide Ceolin
Social Web 2016 Lora Aroyo and Davide Ceolin
Social Web 2016 Lora Aroyo and Davide Ceolin
httpwwwwiredcommagazine201111mf_artsyall1
Social Web 2016 Lora Aroyo and Davide Ceolin
Collaborative Filteringbull Memory-based User-Item matrix ratingspreferences of users =gt
compute similarity between users amp recommend items of similar users
bull Model-based Item-Item matrix similarity (eg based on user ratings) between items =gt recommend items that are similar to the ones the user likes
bull Model-based Clustering cluster users according to their preferences =gt recommend items of users that belong to the same cluster
bull Model-based Bayesian networks P(u likes item B | u likes item A) = how likely is it that a user who likes item A will like item B learn probabilities from user ratingspreferences
bull Others rule-based other data mining techniques
Social Web 2016 Lora Aroyo and Davide Ceolin
bull complete input data is required
bull pre-computation not possible
bull does not scale well bull high quality of
recommendations
bull abstraction (model) of input data
bull pre-computation (partially) possible (model has to be re-built from time to time)
bull scales betterbull abstraction may reduce
recommendation quality
Memory vs Model-based
Social Web 2016 Lora Aroyo and Davide Ceolin
bull collaborative filtering lsquoneighborhoodsrsquo of people with similar interest amp recommending items based on likings in neighborhood
bull limitations next to lsquocold startrsquo and lsquosparsityrsquo the lack of control (over onersquos neighborhood) is also a problem ie cannot add lsquotrustedrsquo people nor exclude lsquostrangersquo ones
bull therefore interest in lsquosocial recommendersrsquo where presence of social connections defines the similarity in interests (eg social tagging CiteULike)bull does a social connection indicate user interest similaritybull how much users interest similarity depends on the strength
of their connectionbull is it feasible to use a social network as a personalized
recommendation[Lin amp Brusilovsky Social Networks and Interest Similarity The Case of CiteULike HTrsquo10]
Social Networks amp Interest Similarity
bull unilaterally connected pairs have more common itemsmetadatatags than non-connected pairsbull highest similarity for direct connections - decreasing with the increase of distance between users in SN bull reciprocal relationship users - significantly larger similarity than users in a unidirectional relationship bull traditional item-level similarity may be less reliable to find similar users in social bookmarking systemsbull peers connected by self-defined social connections could be a useful source for cross-recommendation
Conclusions
Social Web 2016 Lora Aroyo and Davide Ceolin
bull Input characteristics of items amp interests of a user into characteristics of items =gt Recommend items that feature characteristics which meet the userrsquos interests
bull Techniquesbull Data mining methods Cluster items based on their
characteristics =gt Infer usersrsquo interests into clustersbull IR methods Represent items amp users as term vectors
=gt Compute similarity between user profile vector and items
bull Utility-based methods Utility function that gets an item as input the parameters of the utility function are customized via preferences of a user
Content-based Recommendations
Social Web 2016 Lora Aroyo and Davide Ceolin
based on slides from Fabien Abel
Content Features
based on slides from Fabien Abel
User Model
based on slides from Fabien Abel
Recommendations
RecSys Issuesbull Cold-start problem (new user problem) nolittle data available to infer preferences of new
usersbull Changing User Preferences user interests may change over timebull Sparsity problem (new item problem) item descriptions are sparse eg not many user
rated or tagged an itembull Lack of Diversity (overfitting) when adapting too strongly to the preferences of users they
might see samesimilar recommendationsbull Use the right context users do things which might not be relevant for their user model
eg try out things do stuff for other peoplebull Research challenge right balance between serendipity amp personalizationbull Research challenge right way to use the influence of recommendations on userrsquos behavior
Social Web 2016 Lora Aroyo and Davide Ceolin
Announcementsbull Next deadlinesbull Tuesday March 1st 2359
Assignment 2bull Friday March 4th 1000 Post
questionbull Friday March 4th 1700 Vote
questions
image source httpwwwflickrcomphotosbionicteaching1375254387
Hands-on Teaser
bull Your Facebook Friendsrsquo popularity in a spread sheet
bull Locations of your Facebook Friendsbull Tag Cloud of your wall posts
- Social Web 2016
- Slide 2
- Slide 3
- Slide 4
- Slide 5
- Slide 6
- Slide 7
- Slide 8
- Slide 9
- Slide 10
- Slide 11
- Slide 12
- Slide 13
- Slide 14
- Slide 15
- Slide 16
- Slide 17
- Slide 18
- Slide 19
- Slide 20
- Slide 21
- Slide 22
- Slide 23
- Slide 24
- Slide 25
- Slide 26
- Slide 27
- Slide 28
- Slide 29
- Slide 30
- Slide 31
- Slide 32
- Slide 33
- Slide 34
- Slide 35
- Slide 36
- Slide 37
- Slide 38
- Slide 39
- Slide 40
- Slide 41
- Slide 42
- Slide 43
- Slide 44
- Slide 45
- Slide 46
- Slide 47
- Slide 48
- Announcements
- Slide 50
-
From the consumer perspective of an adaptive system
From the provider perspective of an adaptive system
Success Perspectives
Social Web 2016 Lora Aroyo and Davide Ceolin
bull User studies askobserve (selected) people whether you did a good job
bull Log analysis Analyze (click) data and infer whether you did a good job
bull Evaluation of user modelingbull measure quality of profiles directly eg measure
overlap with existing (true) profiles or let people judge the quality of the generated user profiles
bull measure quality of application that exploits the user profile eg apply user modeling strategies in a recommender system
Evaluation Strategies
Social Web 2016 Lora Aroyo and Davide Ceolin
Evaluating User Modeling in RecSys
Social Web 2016 Lora Aroyo and Davide Ceolin
Possible Metrics
Social Web 2016 Lora Aroyo and Davide Ceolin
bull The usual IR metricsbull Precision fraction of retrieved items that are relevantbull Recall fraction of relevant items that have been
retrievedbull F-Measure (harmonic) mean of precision and recall
bull Metrics for evaluating recommendation (rankings)bull Mean Reciprocal Rank (MRR) of first relevant itembull Successk probability that relevant item occurs
within the top kbull If a true ranking is given rank correlations bull Precisionk Recallk amp F-Measurek
bull Metrics for evaluating prediction of user preferencesbull MAE = Mean Absolute Errorbull TrueFalse PositivesNegatives
Predict relevantusefulinteresting items for a given user (in a given context) itrsquos often a ranking task
Recommendation Systems
Social Web 2016 Lora Aroyo and Davide Ceolin
March 28 2013
Social Web 2016 Lora Aroyo and Davide Ceolin
Social Web 2016 Lora Aroyo and Davide Ceolin
Social Web 2016 Lora Aroyo and Davide Ceolin
httpwwwwiredcommagazine201111mf_artsyall1
Social Web 2016 Lora Aroyo and Davide Ceolin
Collaborative Filteringbull Memory-based User-Item matrix ratingspreferences of users =gt
compute similarity between users amp recommend items of similar users
bull Model-based Item-Item matrix similarity (eg based on user ratings) between items =gt recommend items that are similar to the ones the user likes
bull Model-based Clustering cluster users according to their preferences =gt recommend items of users that belong to the same cluster
bull Model-based Bayesian networks P(u likes item B | u likes item A) = how likely is it that a user who likes item A will like item B learn probabilities from user ratingspreferences
bull Others rule-based other data mining techniques
Social Web 2016 Lora Aroyo and Davide Ceolin
bull complete input data is required
bull pre-computation not possible
bull does not scale well bull high quality of
recommendations
bull abstraction (model) of input data
bull pre-computation (partially) possible (model has to be re-built from time to time)
bull scales betterbull abstraction may reduce
recommendation quality
Memory vs Model-based
Social Web 2016 Lora Aroyo and Davide Ceolin
bull collaborative filtering lsquoneighborhoodsrsquo of people with similar interest amp recommending items based on likings in neighborhood
bull limitations next to lsquocold startrsquo and lsquosparsityrsquo the lack of control (over onersquos neighborhood) is also a problem ie cannot add lsquotrustedrsquo people nor exclude lsquostrangersquo ones
bull therefore interest in lsquosocial recommendersrsquo where presence of social connections defines the similarity in interests (eg social tagging CiteULike)bull does a social connection indicate user interest similaritybull how much users interest similarity depends on the strength
of their connectionbull is it feasible to use a social network as a personalized
recommendation[Lin amp Brusilovsky Social Networks and Interest Similarity The Case of CiteULike HTrsquo10]
Social Networks amp Interest Similarity
bull unilaterally connected pairs have more common itemsmetadatatags than non-connected pairsbull highest similarity for direct connections - decreasing with the increase of distance between users in SN bull reciprocal relationship users - significantly larger similarity than users in a unidirectional relationship bull traditional item-level similarity may be less reliable to find similar users in social bookmarking systemsbull peers connected by self-defined social connections could be a useful source for cross-recommendation
Conclusions
Social Web 2016 Lora Aroyo and Davide Ceolin
bull Input characteristics of items amp interests of a user into characteristics of items =gt Recommend items that feature characteristics which meet the userrsquos interests
bull Techniquesbull Data mining methods Cluster items based on their
characteristics =gt Infer usersrsquo interests into clustersbull IR methods Represent items amp users as term vectors
=gt Compute similarity between user profile vector and items
bull Utility-based methods Utility function that gets an item as input the parameters of the utility function are customized via preferences of a user
Content-based Recommendations
Social Web 2016 Lora Aroyo and Davide Ceolin
based on slides from Fabien Abel
Content Features
based on slides from Fabien Abel
User Model
based on slides from Fabien Abel
Recommendations
RecSys Issuesbull Cold-start problem (new user problem) nolittle data available to infer preferences of new
usersbull Changing User Preferences user interests may change over timebull Sparsity problem (new item problem) item descriptions are sparse eg not many user
rated or tagged an itembull Lack of Diversity (overfitting) when adapting too strongly to the preferences of users they
might see samesimilar recommendationsbull Use the right context users do things which might not be relevant for their user model
eg try out things do stuff for other peoplebull Research challenge right balance between serendipity amp personalizationbull Research challenge right way to use the influence of recommendations on userrsquos behavior
Social Web 2016 Lora Aroyo and Davide Ceolin
Announcementsbull Next deadlinesbull Tuesday March 1st 2359
Assignment 2bull Friday March 4th 1000 Post
questionbull Friday March 4th 1700 Vote
questions
image source httpwwwflickrcomphotosbionicteaching1375254387
Hands-on Teaser
bull Your Facebook Friendsrsquo popularity in a spread sheet
bull Locations of your Facebook Friendsbull Tag Cloud of your wall posts
- Social Web 2016
- Slide 2
- Slide 3
- Slide 4
- Slide 5
- Slide 6
- Slide 7
- Slide 8
- Slide 9
- Slide 10
- Slide 11
- Slide 12
- Slide 13
- Slide 14
- Slide 15
- Slide 16
- Slide 17
- Slide 18
- Slide 19
- Slide 20
- Slide 21
- Slide 22
- Slide 23
- Slide 24
- Slide 25
- Slide 26
- Slide 27
- Slide 28
- Slide 29
- Slide 30
- Slide 31
- Slide 32
- Slide 33
- Slide 34
- Slide 35
- Slide 36
- Slide 37
- Slide 38
- Slide 39
- Slide 40
- Slide 41
- Slide 42
- Slide 43
- Slide 44
- Slide 45
- Slide 46
- Slide 47
- Slide 48
- Announcements
- Slide 50
-
bull User studies askobserve (selected) people whether you did a good job
bull Log analysis Analyze (click) data and infer whether you did a good job
bull Evaluation of user modelingbull measure quality of profiles directly eg measure
overlap with existing (true) profiles or let people judge the quality of the generated user profiles
bull measure quality of application that exploits the user profile eg apply user modeling strategies in a recommender system
Evaluation Strategies
Social Web 2016 Lora Aroyo and Davide Ceolin
Evaluating User Modeling in RecSys
Social Web 2016 Lora Aroyo and Davide Ceolin
Possible Metrics
Social Web 2016 Lora Aroyo and Davide Ceolin
bull The usual IR metricsbull Precision fraction of retrieved items that are relevantbull Recall fraction of relevant items that have been
retrievedbull F-Measure (harmonic) mean of precision and recall
bull Metrics for evaluating recommendation (rankings)bull Mean Reciprocal Rank (MRR) of first relevant itembull Successk probability that relevant item occurs
within the top kbull If a true ranking is given rank correlations bull Precisionk Recallk amp F-Measurek
bull Metrics for evaluating prediction of user preferencesbull MAE = Mean Absolute Errorbull TrueFalse PositivesNegatives
Predict relevantusefulinteresting items for a given user (in a given context) itrsquos often a ranking task
Recommendation Systems
Social Web 2016 Lora Aroyo and Davide Ceolin
March 28 2013
Social Web 2016 Lora Aroyo and Davide Ceolin
Social Web 2016 Lora Aroyo and Davide Ceolin
Social Web 2016 Lora Aroyo and Davide Ceolin
httpwwwwiredcommagazine201111mf_artsyall1
Social Web 2016 Lora Aroyo and Davide Ceolin
Collaborative Filteringbull Memory-based User-Item matrix ratingspreferences of users =gt
compute similarity between users amp recommend items of similar users
bull Model-based Item-Item matrix similarity (eg based on user ratings) between items =gt recommend items that are similar to the ones the user likes
bull Model-based Clustering cluster users according to their preferences =gt recommend items of users that belong to the same cluster
bull Model-based Bayesian networks P(u likes item B | u likes item A) = how likely is it that a user who likes item A will like item B learn probabilities from user ratingspreferences
bull Others rule-based other data mining techniques
Social Web 2016 Lora Aroyo and Davide Ceolin
bull complete input data is required
bull pre-computation not possible
bull does not scale well bull high quality of
recommendations
bull abstraction (model) of input data
bull pre-computation (partially) possible (model has to be re-built from time to time)
bull scales betterbull abstraction may reduce
recommendation quality
Memory vs Model-based
Social Web 2016 Lora Aroyo and Davide Ceolin
bull collaborative filtering lsquoneighborhoodsrsquo of people with similar interest amp recommending items based on likings in neighborhood
bull limitations next to lsquocold startrsquo and lsquosparsityrsquo the lack of control (over onersquos neighborhood) is also a problem ie cannot add lsquotrustedrsquo people nor exclude lsquostrangersquo ones
bull therefore interest in lsquosocial recommendersrsquo where presence of social connections defines the similarity in interests (eg social tagging CiteULike)bull does a social connection indicate user interest similaritybull how much users interest similarity depends on the strength
of their connectionbull is it feasible to use a social network as a personalized
recommendation[Lin amp Brusilovsky Social Networks and Interest Similarity The Case of CiteULike HTrsquo10]
Social Networks amp Interest Similarity
bull unilaterally connected pairs have more common itemsmetadatatags than non-connected pairsbull highest similarity for direct connections - decreasing with the increase of distance between users in SN bull reciprocal relationship users - significantly larger similarity than users in a unidirectional relationship bull traditional item-level similarity may be less reliable to find similar users in social bookmarking systemsbull peers connected by self-defined social connections could be a useful source for cross-recommendation
Conclusions
Social Web 2016 Lora Aroyo and Davide Ceolin
bull Input characteristics of items amp interests of a user into characteristics of items =gt Recommend items that feature characteristics which meet the userrsquos interests
bull Techniquesbull Data mining methods Cluster items based on their
characteristics =gt Infer usersrsquo interests into clustersbull IR methods Represent items amp users as term vectors
=gt Compute similarity between user profile vector and items
bull Utility-based methods Utility function that gets an item as input the parameters of the utility function are customized via preferences of a user
Content-based Recommendations
Social Web 2016 Lora Aroyo and Davide Ceolin
based on slides from Fabien Abel
Content Features
based on slides from Fabien Abel
User Model
based on slides from Fabien Abel
Recommendations
RecSys Issuesbull Cold-start problem (new user problem) nolittle data available to infer preferences of new
usersbull Changing User Preferences user interests may change over timebull Sparsity problem (new item problem) item descriptions are sparse eg not many user
rated or tagged an itembull Lack of Diversity (overfitting) when adapting too strongly to the preferences of users they
might see samesimilar recommendationsbull Use the right context users do things which might not be relevant for their user model
eg try out things do stuff for other peoplebull Research challenge right balance between serendipity amp personalizationbull Research challenge right way to use the influence of recommendations on userrsquos behavior
Social Web 2016 Lora Aroyo and Davide Ceolin
Announcementsbull Next deadlinesbull Tuesday March 1st 2359
Assignment 2bull Friday March 4th 1000 Post
questionbull Friday March 4th 1700 Vote
questions
image source httpwwwflickrcomphotosbionicteaching1375254387
Hands-on Teaser
bull Your Facebook Friendsrsquo popularity in a spread sheet
bull Locations of your Facebook Friendsbull Tag Cloud of your wall posts
- Social Web 2016
- Slide 2
- Slide 3
- Slide 4
- Slide 5
- Slide 6
- Slide 7
- Slide 8
- Slide 9
- Slide 10
- Slide 11
- Slide 12
- Slide 13
- Slide 14
- Slide 15
- Slide 16
- Slide 17
- Slide 18
- Slide 19
- Slide 20
- Slide 21
- Slide 22
- Slide 23
- Slide 24
- Slide 25
- Slide 26
- Slide 27
- Slide 28
- Slide 29
- Slide 30
- Slide 31
- Slide 32
- Slide 33
- Slide 34
- Slide 35
- Slide 36
- Slide 37
- Slide 38
- Slide 39
- Slide 40
- Slide 41
- Slide 42
- Slide 43
- Slide 44
- Slide 45
- Slide 46
- Slide 47
- Slide 48
- Announcements
- Slide 50
-
Evaluating User Modeling in RecSys
Social Web 2016 Lora Aroyo and Davide Ceolin
Possible Metrics
Social Web 2016 Lora Aroyo and Davide Ceolin
bull The usual IR metricsbull Precision fraction of retrieved items that are relevantbull Recall fraction of relevant items that have been
retrievedbull F-Measure (harmonic) mean of precision and recall
bull Metrics for evaluating recommendation (rankings)bull Mean Reciprocal Rank (MRR) of first relevant itembull Successk probability that relevant item occurs
within the top kbull If a true ranking is given rank correlations bull Precisionk Recallk amp F-Measurek
bull Metrics for evaluating prediction of user preferencesbull MAE = Mean Absolute Errorbull TrueFalse PositivesNegatives
Predict relevantusefulinteresting items for a given user (in a given context) itrsquos often a ranking task
Recommendation Systems
Social Web 2016 Lora Aroyo and Davide Ceolin
March 28 2013
Social Web 2016 Lora Aroyo and Davide Ceolin
Social Web 2016 Lora Aroyo and Davide Ceolin
Social Web 2016 Lora Aroyo and Davide Ceolin
httpwwwwiredcommagazine201111mf_artsyall1
Social Web 2016 Lora Aroyo and Davide Ceolin
Collaborative Filteringbull Memory-based User-Item matrix ratingspreferences of users =gt
compute similarity between users amp recommend items of similar users
bull Model-based Item-Item matrix similarity (eg based on user ratings) between items =gt recommend items that are similar to the ones the user likes
bull Model-based Clustering cluster users according to their preferences =gt recommend items of users that belong to the same cluster
bull Model-based Bayesian networks P(u likes item B | u likes item A) = how likely is it that a user who likes item A will like item B learn probabilities from user ratingspreferences
bull Others rule-based other data mining techniques
Social Web 2016 Lora Aroyo and Davide Ceolin
bull complete input data is required
bull pre-computation not possible
bull does not scale well bull high quality of
recommendations
bull abstraction (model) of input data
bull pre-computation (partially) possible (model has to be re-built from time to time)
bull scales betterbull abstraction may reduce
recommendation quality
Memory vs Model-based
Social Web 2016 Lora Aroyo and Davide Ceolin
bull collaborative filtering lsquoneighborhoodsrsquo of people with similar interest amp recommending items based on likings in neighborhood
bull limitations next to lsquocold startrsquo and lsquosparsityrsquo the lack of control (over onersquos neighborhood) is also a problem ie cannot add lsquotrustedrsquo people nor exclude lsquostrangersquo ones
bull therefore interest in lsquosocial recommendersrsquo where presence of social connections defines the similarity in interests (eg social tagging CiteULike)bull does a social connection indicate user interest similaritybull how much users interest similarity depends on the strength
of their connectionbull is it feasible to use a social network as a personalized
recommendation[Lin amp Brusilovsky Social Networks and Interest Similarity The Case of CiteULike HTrsquo10]
Social Networks amp Interest Similarity
bull unilaterally connected pairs have more common itemsmetadatatags than non-connected pairsbull highest similarity for direct connections - decreasing with the increase of distance between users in SN bull reciprocal relationship users - significantly larger similarity than users in a unidirectional relationship bull traditional item-level similarity may be less reliable to find similar users in social bookmarking systemsbull peers connected by self-defined social connections could be a useful source for cross-recommendation
Conclusions
Social Web 2016 Lora Aroyo and Davide Ceolin
bull Input characteristics of items amp interests of a user into characteristics of items =gt Recommend items that feature characteristics which meet the userrsquos interests
bull Techniquesbull Data mining methods Cluster items based on their
characteristics =gt Infer usersrsquo interests into clustersbull IR methods Represent items amp users as term vectors
=gt Compute similarity between user profile vector and items
bull Utility-based methods Utility function that gets an item as input the parameters of the utility function are customized via preferences of a user
Content-based Recommendations
Social Web 2016 Lora Aroyo and Davide Ceolin
based on slides from Fabien Abel
Content Features
based on slides from Fabien Abel
User Model
based on slides from Fabien Abel
Recommendations
RecSys Issuesbull Cold-start problem (new user problem) nolittle data available to infer preferences of new
usersbull Changing User Preferences user interests may change over timebull Sparsity problem (new item problem) item descriptions are sparse eg not many user
rated or tagged an itembull Lack of Diversity (overfitting) when adapting too strongly to the preferences of users they
might see samesimilar recommendationsbull Use the right context users do things which might not be relevant for their user model
eg try out things do stuff for other peoplebull Research challenge right balance between serendipity amp personalizationbull Research challenge right way to use the influence of recommendations on userrsquos behavior
Social Web 2016 Lora Aroyo and Davide Ceolin
Announcementsbull Next deadlinesbull Tuesday March 1st 2359
Assignment 2bull Friday March 4th 1000 Post
questionbull Friday March 4th 1700 Vote
questions
image source httpwwwflickrcomphotosbionicteaching1375254387
Hands-on Teaser
bull Your Facebook Friendsrsquo popularity in a spread sheet
bull Locations of your Facebook Friendsbull Tag Cloud of your wall posts
- Social Web 2016
- Slide 2
- Slide 3
- Slide 4
- Slide 5
- Slide 6
- Slide 7
- Slide 8
- Slide 9
- Slide 10
- Slide 11
- Slide 12
- Slide 13
- Slide 14
- Slide 15
- Slide 16
- Slide 17
- Slide 18
- Slide 19
- Slide 20
- Slide 21
- Slide 22
- Slide 23
- Slide 24
- Slide 25
- Slide 26
- Slide 27
- Slide 28
- Slide 29
- Slide 30
- Slide 31
- Slide 32
- Slide 33
- Slide 34
- Slide 35
- Slide 36
- Slide 37
- Slide 38
- Slide 39
- Slide 40
- Slide 41
- Slide 42
- Slide 43
- Slide 44
- Slide 45
- Slide 46
- Slide 47
- Slide 48
- Announcements
- Slide 50
-
Possible Metrics
Social Web 2016 Lora Aroyo and Davide Ceolin
bull The usual IR metricsbull Precision fraction of retrieved items that are relevantbull Recall fraction of relevant items that have been
retrievedbull F-Measure (harmonic) mean of precision and recall
bull Metrics for evaluating recommendation (rankings)bull Mean Reciprocal Rank (MRR) of first relevant itembull Successk probability that relevant item occurs
within the top kbull If a true ranking is given rank correlations bull Precisionk Recallk amp F-Measurek
bull Metrics for evaluating prediction of user preferencesbull MAE = Mean Absolute Errorbull TrueFalse PositivesNegatives
Predict relevantusefulinteresting items for a given user (in a given context) itrsquos often a ranking task
Recommendation Systems
Social Web 2016 Lora Aroyo and Davide Ceolin
March 28 2013
Social Web 2016 Lora Aroyo and Davide Ceolin
Social Web 2016 Lora Aroyo and Davide Ceolin
Social Web 2016 Lora Aroyo and Davide Ceolin
httpwwwwiredcommagazine201111mf_artsyall1
Social Web 2016 Lora Aroyo and Davide Ceolin
Collaborative Filteringbull Memory-based User-Item matrix ratingspreferences of users =gt
compute similarity between users amp recommend items of similar users
bull Model-based Item-Item matrix similarity (eg based on user ratings) between items =gt recommend items that are similar to the ones the user likes
bull Model-based Clustering cluster users according to their preferences =gt recommend items of users that belong to the same cluster
bull Model-based Bayesian networks P(u likes item B | u likes item A) = how likely is it that a user who likes item A will like item B learn probabilities from user ratingspreferences
bull Others rule-based other data mining techniques
Social Web 2016 Lora Aroyo and Davide Ceolin
bull complete input data is required
bull pre-computation not possible
bull does not scale well bull high quality of
recommendations
bull abstraction (model) of input data
bull pre-computation (partially) possible (model has to be re-built from time to time)
bull scales betterbull abstraction may reduce
recommendation quality
Memory vs Model-based
Social Web 2016 Lora Aroyo and Davide Ceolin
bull collaborative filtering lsquoneighborhoodsrsquo of people with similar interest amp recommending items based on likings in neighborhood
bull limitations next to lsquocold startrsquo and lsquosparsityrsquo the lack of control (over onersquos neighborhood) is also a problem ie cannot add lsquotrustedrsquo people nor exclude lsquostrangersquo ones
bull therefore interest in lsquosocial recommendersrsquo where presence of social connections defines the similarity in interests (eg social tagging CiteULike)bull does a social connection indicate user interest similaritybull how much users interest similarity depends on the strength
of their connectionbull is it feasible to use a social network as a personalized
recommendation[Lin amp Brusilovsky Social Networks and Interest Similarity The Case of CiteULike HTrsquo10]
Social Networks amp Interest Similarity
bull unilaterally connected pairs have more common itemsmetadatatags than non-connected pairsbull highest similarity for direct connections - decreasing with the increase of distance between users in SN bull reciprocal relationship users - significantly larger similarity than users in a unidirectional relationship bull traditional item-level similarity may be less reliable to find similar users in social bookmarking systemsbull peers connected by self-defined social connections could be a useful source for cross-recommendation
Conclusions
Social Web 2016 Lora Aroyo and Davide Ceolin
bull Input characteristics of items amp interests of a user into characteristics of items =gt Recommend items that feature characteristics which meet the userrsquos interests
bull Techniquesbull Data mining methods Cluster items based on their
characteristics =gt Infer usersrsquo interests into clustersbull IR methods Represent items amp users as term vectors
=gt Compute similarity between user profile vector and items
bull Utility-based methods Utility function that gets an item as input the parameters of the utility function are customized via preferences of a user
Content-based Recommendations
Social Web 2016 Lora Aroyo and Davide Ceolin
based on slides from Fabien Abel
Content Features
based on slides from Fabien Abel
User Model
based on slides from Fabien Abel
Recommendations
RecSys Issuesbull Cold-start problem (new user problem) nolittle data available to infer preferences of new
usersbull Changing User Preferences user interests may change over timebull Sparsity problem (new item problem) item descriptions are sparse eg not many user
rated or tagged an itembull Lack of Diversity (overfitting) when adapting too strongly to the preferences of users they
might see samesimilar recommendationsbull Use the right context users do things which might not be relevant for their user model
eg try out things do stuff for other peoplebull Research challenge right balance between serendipity amp personalizationbull Research challenge right way to use the influence of recommendations on userrsquos behavior
Social Web 2016 Lora Aroyo and Davide Ceolin
Announcementsbull Next deadlinesbull Tuesday March 1st 2359
Assignment 2bull Friday March 4th 1000 Post
questionbull Friday March 4th 1700 Vote
questions
image source httpwwwflickrcomphotosbionicteaching1375254387
Hands-on Teaser
bull Your Facebook Friendsrsquo popularity in a spread sheet
bull Locations of your Facebook Friendsbull Tag Cloud of your wall posts
- Social Web 2016
- Slide 2
- Slide 3
- Slide 4
- Slide 5
- Slide 6
- Slide 7
- Slide 8
- Slide 9
- Slide 10
- Slide 11
- Slide 12
- Slide 13
- Slide 14
- Slide 15
- Slide 16
- Slide 17
- Slide 18
- Slide 19
- Slide 20
- Slide 21
- Slide 22
- Slide 23
- Slide 24
- Slide 25
- Slide 26
- Slide 27
- Slide 28
- Slide 29
- Slide 30
- Slide 31
- Slide 32
- Slide 33
- Slide 34
- Slide 35
- Slide 36
- Slide 37
- Slide 38
- Slide 39
- Slide 40
- Slide 41
- Slide 42
- Slide 43
- Slide 44
- Slide 45
- Slide 46
- Slide 47
- Slide 48
- Announcements
- Slide 50
-
Predict relevantusefulinteresting items for a given user (in a given context) itrsquos often a ranking task
Recommendation Systems
Social Web 2016 Lora Aroyo and Davide Ceolin
March 28 2013
Social Web 2016 Lora Aroyo and Davide Ceolin
Social Web 2016 Lora Aroyo and Davide Ceolin
Social Web 2016 Lora Aroyo and Davide Ceolin
httpwwwwiredcommagazine201111mf_artsyall1
Social Web 2016 Lora Aroyo and Davide Ceolin
Collaborative Filteringbull Memory-based User-Item matrix ratingspreferences of users =gt
compute similarity between users amp recommend items of similar users
bull Model-based Item-Item matrix similarity (eg based on user ratings) between items =gt recommend items that are similar to the ones the user likes
bull Model-based Clustering cluster users according to their preferences =gt recommend items of users that belong to the same cluster
bull Model-based Bayesian networks P(u likes item B | u likes item A) = how likely is it that a user who likes item A will like item B learn probabilities from user ratingspreferences
bull Others rule-based other data mining techniques
Social Web 2016 Lora Aroyo and Davide Ceolin
bull complete input data is required
bull pre-computation not possible
bull does not scale well bull high quality of
recommendations
bull abstraction (model) of input data
bull pre-computation (partially) possible (model has to be re-built from time to time)
bull scales betterbull abstraction may reduce
recommendation quality
Memory vs Model-based
Social Web 2016 Lora Aroyo and Davide Ceolin
bull collaborative filtering lsquoneighborhoodsrsquo of people with similar interest amp recommending items based on likings in neighborhood
bull limitations next to lsquocold startrsquo and lsquosparsityrsquo the lack of control (over onersquos neighborhood) is also a problem ie cannot add lsquotrustedrsquo people nor exclude lsquostrangersquo ones
bull therefore interest in lsquosocial recommendersrsquo where presence of social connections defines the similarity in interests (eg social tagging CiteULike)bull does a social connection indicate user interest similaritybull how much users interest similarity depends on the strength
of their connectionbull is it feasible to use a social network as a personalized
recommendation[Lin amp Brusilovsky Social Networks and Interest Similarity The Case of CiteULike HTrsquo10]
Social Networks amp Interest Similarity
bull unilaterally connected pairs have more common itemsmetadatatags than non-connected pairsbull highest similarity for direct connections - decreasing with the increase of distance between users in SN bull reciprocal relationship users - significantly larger similarity than users in a unidirectional relationship bull traditional item-level similarity may be less reliable to find similar users in social bookmarking systemsbull peers connected by self-defined social connections could be a useful source for cross-recommendation
Conclusions
Social Web 2016 Lora Aroyo and Davide Ceolin
bull Input characteristics of items amp interests of a user into characteristics of items =gt Recommend items that feature characteristics which meet the userrsquos interests
bull Techniquesbull Data mining methods Cluster items based on their
characteristics =gt Infer usersrsquo interests into clustersbull IR methods Represent items amp users as term vectors
=gt Compute similarity between user profile vector and items
bull Utility-based methods Utility function that gets an item as input the parameters of the utility function are customized via preferences of a user
Content-based Recommendations
Social Web 2016 Lora Aroyo and Davide Ceolin
based on slides from Fabien Abel
Content Features
based on slides from Fabien Abel
User Model
based on slides from Fabien Abel
Recommendations
RecSys Issuesbull Cold-start problem (new user problem) nolittle data available to infer preferences of new
usersbull Changing User Preferences user interests may change over timebull Sparsity problem (new item problem) item descriptions are sparse eg not many user
rated or tagged an itembull Lack of Diversity (overfitting) when adapting too strongly to the preferences of users they
might see samesimilar recommendationsbull Use the right context users do things which might not be relevant for their user model
eg try out things do stuff for other peoplebull Research challenge right balance between serendipity amp personalizationbull Research challenge right way to use the influence of recommendations on userrsquos behavior
Social Web 2016 Lora Aroyo and Davide Ceolin
Announcementsbull Next deadlinesbull Tuesday March 1st 2359
Assignment 2bull Friday March 4th 1000 Post
questionbull Friday March 4th 1700 Vote
questions
image source httpwwwflickrcomphotosbionicteaching1375254387
Hands-on Teaser
bull Your Facebook Friendsrsquo popularity in a spread sheet
bull Locations of your Facebook Friendsbull Tag Cloud of your wall posts
- Social Web 2016
- Slide 2
- Slide 3
- Slide 4
- Slide 5
- Slide 6
- Slide 7
- Slide 8
- Slide 9
- Slide 10
- Slide 11
- Slide 12
- Slide 13
- Slide 14
- Slide 15
- Slide 16
- Slide 17
- Slide 18
- Slide 19
- Slide 20
- Slide 21
- Slide 22
- Slide 23
- Slide 24
- Slide 25
- Slide 26
- Slide 27
- Slide 28
- Slide 29
- Slide 30
- Slide 31
- Slide 32
- Slide 33
- Slide 34
- Slide 35
- Slide 36
- Slide 37
- Slide 38
- Slide 39
- Slide 40
- Slide 41
- Slide 42
- Slide 43
- Slide 44
- Slide 45
- Slide 46
- Slide 47
- Slide 48
- Announcements
- Slide 50
-
Social Web 2016 Lora Aroyo and Davide Ceolin
March 28 2013
Social Web 2016 Lora Aroyo and Davide Ceolin
Social Web 2016 Lora Aroyo and Davide Ceolin
Social Web 2016 Lora Aroyo and Davide Ceolin
httpwwwwiredcommagazine201111mf_artsyall1
Social Web 2016 Lora Aroyo and Davide Ceolin
Collaborative Filteringbull Memory-based User-Item matrix ratingspreferences of users =gt
compute similarity between users amp recommend items of similar users
bull Model-based Item-Item matrix similarity (eg based on user ratings) between items =gt recommend items that are similar to the ones the user likes
bull Model-based Clustering cluster users according to their preferences =gt recommend items of users that belong to the same cluster
bull Model-based Bayesian networks P(u likes item B | u likes item A) = how likely is it that a user who likes item A will like item B learn probabilities from user ratingspreferences
bull Others rule-based other data mining techniques
Social Web 2016 Lora Aroyo and Davide Ceolin
bull complete input data is required
bull pre-computation not possible
bull does not scale well bull high quality of
recommendations
bull abstraction (model) of input data
bull pre-computation (partially) possible (model has to be re-built from time to time)
bull scales betterbull abstraction may reduce
recommendation quality
Memory vs Model-based
Social Web 2016 Lora Aroyo and Davide Ceolin
bull collaborative filtering lsquoneighborhoodsrsquo of people with similar interest amp recommending items based on likings in neighborhood
bull limitations next to lsquocold startrsquo and lsquosparsityrsquo the lack of control (over onersquos neighborhood) is also a problem ie cannot add lsquotrustedrsquo people nor exclude lsquostrangersquo ones
bull therefore interest in lsquosocial recommendersrsquo where presence of social connections defines the similarity in interests (eg social tagging CiteULike)bull does a social connection indicate user interest similaritybull how much users interest similarity depends on the strength
of their connectionbull is it feasible to use a social network as a personalized
recommendation[Lin amp Brusilovsky Social Networks and Interest Similarity The Case of CiteULike HTrsquo10]
Social Networks amp Interest Similarity
bull unilaterally connected pairs have more common itemsmetadatatags than non-connected pairsbull highest similarity for direct connections - decreasing with the increase of distance between users in SN bull reciprocal relationship users - significantly larger similarity than users in a unidirectional relationship bull traditional item-level similarity may be less reliable to find similar users in social bookmarking systemsbull peers connected by self-defined social connections could be a useful source for cross-recommendation
Conclusions
Social Web 2016 Lora Aroyo and Davide Ceolin
bull Input characteristics of items amp interests of a user into characteristics of items =gt Recommend items that feature characteristics which meet the userrsquos interests
bull Techniquesbull Data mining methods Cluster items based on their
characteristics =gt Infer usersrsquo interests into clustersbull IR methods Represent items amp users as term vectors
=gt Compute similarity between user profile vector and items
bull Utility-based methods Utility function that gets an item as input the parameters of the utility function are customized via preferences of a user
Content-based Recommendations
Social Web 2016 Lora Aroyo and Davide Ceolin
based on slides from Fabien Abel
Content Features
based on slides from Fabien Abel
User Model
based on slides from Fabien Abel
Recommendations
RecSys Issuesbull Cold-start problem (new user problem) nolittle data available to infer preferences of new
usersbull Changing User Preferences user interests may change over timebull Sparsity problem (new item problem) item descriptions are sparse eg not many user
rated or tagged an itembull Lack of Diversity (overfitting) when adapting too strongly to the preferences of users they
might see samesimilar recommendationsbull Use the right context users do things which might not be relevant for their user model
eg try out things do stuff for other peoplebull Research challenge right balance between serendipity amp personalizationbull Research challenge right way to use the influence of recommendations on userrsquos behavior
Social Web 2016 Lora Aroyo and Davide Ceolin
Announcementsbull Next deadlinesbull Tuesday March 1st 2359
Assignment 2bull Friday March 4th 1000 Post
questionbull Friday March 4th 1700 Vote
questions
image source httpwwwflickrcomphotosbionicteaching1375254387
Hands-on Teaser
bull Your Facebook Friendsrsquo popularity in a spread sheet
bull Locations of your Facebook Friendsbull Tag Cloud of your wall posts
- Social Web 2016
- Slide 2
- Slide 3
- Slide 4
- Slide 5
- Slide 6
- Slide 7
- Slide 8
- Slide 9
- Slide 10
- Slide 11
- Slide 12
- Slide 13
- Slide 14
- Slide 15
- Slide 16
- Slide 17
- Slide 18
- Slide 19
- Slide 20
- Slide 21
- Slide 22
- Slide 23
- Slide 24
- Slide 25
- Slide 26
- Slide 27
- Slide 28
- Slide 29
- Slide 30
- Slide 31
- Slide 32
- Slide 33
- Slide 34
- Slide 35
- Slide 36
- Slide 37
- Slide 38
- Slide 39
- Slide 40
- Slide 41
- Slide 42
- Slide 43
- Slide 44
- Slide 45
- Slide 46
- Slide 47
- Slide 48
- Announcements
- Slide 50
-
March 28 2013
Social Web 2016 Lora Aroyo and Davide Ceolin
Social Web 2016 Lora Aroyo and Davide Ceolin
Social Web 2016 Lora Aroyo and Davide Ceolin
httpwwwwiredcommagazine201111mf_artsyall1
Social Web 2016 Lora Aroyo and Davide Ceolin
Collaborative Filteringbull Memory-based User-Item matrix ratingspreferences of users =gt
compute similarity between users amp recommend items of similar users
bull Model-based Item-Item matrix similarity (eg based on user ratings) between items =gt recommend items that are similar to the ones the user likes
bull Model-based Clustering cluster users according to their preferences =gt recommend items of users that belong to the same cluster
bull Model-based Bayesian networks P(u likes item B | u likes item A) = how likely is it that a user who likes item A will like item B learn probabilities from user ratingspreferences
bull Others rule-based other data mining techniques
Social Web 2016 Lora Aroyo and Davide Ceolin
bull complete input data is required
bull pre-computation not possible
bull does not scale well bull high quality of
recommendations
bull abstraction (model) of input data
bull pre-computation (partially) possible (model has to be re-built from time to time)
bull scales betterbull abstraction may reduce
recommendation quality
Memory vs Model-based
Social Web 2016 Lora Aroyo and Davide Ceolin
bull collaborative filtering lsquoneighborhoodsrsquo of people with similar interest amp recommending items based on likings in neighborhood
bull limitations next to lsquocold startrsquo and lsquosparsityrsquo the lack of control (over onersquos neighborhood) is also a problem ie cannot add lsquotrustedrsquo people nor exclude lsquostrangersquo ones
bull therefore interest in lsquosocial recommendersrsquo where presence of social connections defines the similarity in interests (eg social tagging CiteULike)bull does a social connection indicate user interest similaritybull how much users interest similarity depends on the strength
of their connectionbull is it feasible to use a social network as a personalized
recommendation[Lin amp Brusilovsky Social Networks and Interest Similarity The Case of CiteULike HTrsquo10]
Social Networks amp Interest Similarity
bull unilaterally connected pairs have more common itemsmetadatatags than non-connected pairsbull highest similarity for direct connections - decreasing with the increase of distance between users in SN bull reciprocal relationship users - significantly larger similarity than users in a unidirectional relationship bull traditional item-level similarity may be less reliable to find similar users in social bookmarking systemsbull peers connected by self-defined social connections could be a useful source for cross-recommendation
Conclusions
Social Web 2016 Lora Aroyo and Davide Ceolin
bull Input characteristics of items amp interests of a user into characteristics of items =gt Recommend items that feature characteristics which meet the userrsquos interests
bull Techniquesbull Data mining methods Cluster items based on their
characteristics =gt Infer usersrsquo interests into clustersbull IR methods Represent items amp users as term vectors
=gt Compute similarity between user profile vector and items
bull Utility-based methods Utility function that gets an item as input the parameters of the utility function are customized via preferences of a user
Content-based Recommendations
Social Web 2016 Lora Aroyo and Davide Ceolin
based on slides from Fabien Abel
Content Features
based on slides from Fabien Abel
User Model
based on slides from Fabien Abel
Recommendations
RecSys Issuesbull Cold-start problem (new user problem) nolittle data available to infer preferences of new
usersbull Changing User Preferences user interests may change over timebull Sparsity problem (new item problem) item descriptions are sparse eg not many user
rated or tagged an itembull Lack of Diversity (overfitting) when adapting too strongly to the preferences of users they
might see samesimilar recommendationsbull Use the right context users do things which might not be relevant for their user model
eg try out things do stuff for other peoplebull Research challenge right balance between serendipity amp personalizationbull Research challenge right way to use the influence of recommendations on userrsquos behavior
Social Web 2016 Lora Aroyo and Davide Ceolin
Announcementsbull Next deadlinesbull Tuesday March 1st 2359
Assignment 2bull Friday March 4th 1000 Post
questionbull Friday March 4th 1700 Vote
questions
image source httpwwwflickrcomphotosbionicteaching1375254387
Hands-on Teaser
bull Your Facebook Friendsrsquo popularity in a spread sheet
bull Locations of your Facebook Friendsbull Tag Cloud of your wall posts
- Social Web 2016
- Slide 2
- Slide 3
- Slide 4
- Slide 5
- Slide 6
- Slide 7
- Slide 8
- Slide 9
- Slide 10
- Slide 11
- Slide 12
- Slide 13
- Slide 14
- Slide 15
- Slide 16
- Slide 17
- Slide 18
- Slide 19
- Slide 20
- Slide 21
- Slide 22
- Slide 23
- Slide 24
- Slide 25
- Slide 26
- Slide 27
- Slide 28
- Slide 29
- Slide 30
- Slide 31
- Slide 32
- Slide 33
- Slide 34
- Slide 35
- Slide 36
- Slide 37
- Slide 38
- Slide 39
- Slide 40
- Slide 41
- Slide 42
- Slide 43
- Slide 44
- Slide 45
- Slide 46
- Slide 47
- Slide 48
- Announcements
- Slide 50
-
Social Web 2016 Lora Aroyo and Davide Ceolin
Social Web 2016 Lora Aroyo and Davide Ceolin
httpwwwwiredcommagazine201111mf_artsyall1
Social Web 2016 Lora Aroyo and Davide Ceolin
Collaborative Filteringbull Memory-based User-Item matrix ratingspreferences of users =gt
compute similarity between users amp recommend items of similar users
bull Model-based Item-Item matrix similarity (eg based on user ratings) between items =gt recommend items that are similar to the ones the user likes
bull Model-based Clustering cluster users according to their preferences =gt recommend items of users that belong to the same cluster
bull Model-based Bayesian networks P(u likes item B | u likes item A) = how likely is it that a user who likes item A will like item B learn probabilities from user ratingspreferences
bull Others rule-based other data mining techniques
Social Web 2016 Lora Aroyo and Davide Ceolin
bull complete input data is required
bull pre-computation not possible
bull does not scale well bull high quality of
recommendations
bull abstraction (model) of input data
bull pre-computation (partially) possible (model has to be re-built from time to time)
bull scales betterbull abstraction may reduce
recommendation quality
Memory vs Model-based
Social Web 2016 Lora Aroyo and Davide Ceolin
bull collaborative filtering lsquoneighborhoodsrsquo of people with similar interest amp recommending items based on likings in neighborhood
bull limitations next to lsquocold startrsquo and lsquosparsityrsquo the lack of control (over onersquos neighborhood) is also a problem ie cannot add lsquotrustedrsquo people nor exclude lsquostrangersquo ones
bull therefore interest in lsquosocial recommendersrsquo where presence of social connections defines the similarity in interests (eg social tagging CiteULike)bull does a social connection indicate user interest similaritybull how much users interest similarity depends on the strength
of their connectionbull is it feasible to use a social network as a personalized
recommendation[Lin amp Brusilovsky Social Networks and Interest Similarity The Case of CiteULike HTrsquo10]
Social Networks amp Interest Similarity
bull unilaterally connected pairs have more common itemsmetadatatags than non-connected pairsbull highest similarity for direct connections - decreasing with the increase of distance between users in SN bull reciprocal relationship users - significantly larger similarity than users in a unidirectional relationship bull traditional item-level similarity may be less reliable to find similar users in social bookmarking systemsbull peers connected by self-defined social connections could be a useful source for cross-recommendation
Conclusions
Social Web 2016 Lora Aroyo and Davide Ceolin
bull Input characteristics of items amp interests of a user into characteristics of items =gt Recommend items that feature characteristics which meet the userrsquos interests
bull Techniquesbull Data mining methods Cluster items based on their
characteristics =gt Infer usersrsquo interests into clustersbull IR methods Represent items amp users as term vectors
=gt Compute similarity between user profile vector and items
bull Utility-based methods Utility function that gets an item as input the parameters of the utility function are customized via preferences of a user
Content-based Recommendations
Social Web 2016 Lora Aroyo and Davide Ceolin
based on slides from Fabien Abel
Content Features
based on slides from Fabien Abel
User Model
based on slides from Fabien Abel
Recommendations
RecSys Issuesbull Cold-start problem (new user problem) nolittle data available to infer preferences of new
usersbull Changing User Preferences user interests may change over timebull Sparsity problem (new item problem) item descriptions are sparse eg not many user
rated or tagged an itembull Lack of Diversity (overfitting) when adapting too strongly to the preferences of users they
might see samesimilar recommendationsbull Use the right context users do things which might not be relevant for their user model
eg try out things do stuff for other peoplebull Research challenge right balance between serendipity amp personalizationbull Research challenge right way to use the influence of recommendations on userrsquos behavior
Social Web 2016 Lora Aroyo and Davide Ceolin
Announcementsbull Next deadlinesbull Tuesday March 1st 2359
Assignment 2bull Friday March 4th 1000 Post
questionbull Friday March 4th 1700 Vote
questions
image source httpwwwflickrcomphotosbionicteaching1375254387
Hands-on Teaser
bull Your Facebook Friendsrsquo popularity in a spread sheet
bull Locations of your Facebook Friendsbull Tag Cloud of your wall posts
- Social Web 2016
- Slide 2
- Slide 3
- Slide 4
- Slide 5
- Slide 6
- Slide 7
- Slide 8
- Slide 9
- Slide 10
- Slide 11
- Slide 12
- Slide 13
- Slide 14
- Slide 15
- Slide 16
- Slide 17
- Slide 18
- Slide 19
- Slide 20
- Slide 21
- Slide 22
- Slide 23
- Slide 24
- Slide 25
- Slide 26
- Slide 27
- Slide 28
- Slide 29
- Slide 30
- Slide 31
- Slide 32
- Slide 33
- Slide 34
- Slide 35
- Slide 36
- Slide 37
- Slide 38
- Slide 39
- Slide 40
- Slide 41
- Slide 42
- Slide 43
- Slide 44
- Slide 45
- Slide 46
- Slide 47
- Slide 48
- Announcements
- Slide 50
-
Social Web 2016 Lora Aroyo and Davide Ceolin
httpwwwwiredcommagazine201111mf_artsyall1
Social Web 2016 Lora Aroyo and Davide Ceolin
Collaborative Filteringbull Memory-based User-Item matrix ratingspreferences of users =gt
compute similarity between users amp recommend items of similar users
bull Model-based Item-Item matrix similarity (eg based on user ratings) between items =gt recommend items that are similar to the ones the user likes
bull Model-based Clustering cluster users according to their preferences =gt recommend items of users that belong to the same cluster
bull Model-based Bayesian networks P(u likes item B | u likes item A) = how likely is it that a user who likes item A will like item B learn probabilities from user ratingspreferences
bull Others rule-based other data mining techniques
Social Web 2016 Lora Aroyo and Davide Ceolin
bull complete input data is required
bull pre-computation not possible
bull does not scale well bull high quality of
recommendations
bull abstraction (model) of input data
bull pre-computation (partially) possible (model has to be re-built from time to time)
bull scales betterbull abstraction may reduce
recommendation quality
Memory vs Model-based
Social Web 2016 Lora Aroyo and Davide Ceolin
bull collaborative filtering lsquoneighborhoodsrsquo of people with similar interest amp recommending items based on likings in neighborhood
bull limitations next to lsquocold startrsquo and lsquosparsityrsquo the lack of control (over onersquos neighborhood) is also a problem ie cannot add lsquotrustedrsquo people nor exclude lsquostrangersquo ones
bull therefore interest in lsquosocial recommendersrsquo where presence of social connections defines the similarity in interests (eg social tagging CiteULike)bull does a social connection indicate user interest similaritybull how much users interest similarity depends on the strength
of their connectionbull is it feasible to use a social network as a personalized
recommendation[Lin amp Brusilovsky Social Networks and Interest Similarity The Case of CiteULike HTrsquo10]
Social Networks amp Interest Similarity
bull unilaterally connected pairs have more common itemsmetadatatags than non-connected pairsbull highest similarity for direct connections - decreasing with the increase of distance between users in SN bull reciprocal relationship users - significantly larger similarity than users in a unidirectional relationship bull traditional item-level similarity may be less reliable to find similar users in social bookmarking systemsbull peers connected by self-defined social connections could be a useful source for cross-recommendation
Conclusions
Social Web 2016 Lora Aroyo and Davide Ceolin
bull Input characteristics of items amp interests of a user into characteristics of items =gt Recommend items that feature characteristics which meet the userrsquos interests
bull Techniquesbull Data mining methods Cluster items based on their
characteristics =gt Infer usersrsquo interests into clustersbull IR methods Represent items amp users as term vectors
=gt Compute similarity between user profile vector and items
bull Utility-based methods Utility function that gets an item as input the parameters of the utility function are customized via preferences of a user
Content-based Recommendations
Social Web 2016 Lora Aroyo and Davide Ceolin
based on slides from Fabien Abel
Content Features
based on slides from Fabien Abel
User Model
based on slides from Fabien Abel
Recommendations
RecSys Issuesbull Cold-start problem (new user problem) nolittle data available to infer preferences of new
usersbull Changing User Preferences user interests may change over timebull Sparsity problem (new item problem) item descriptions are sparse eg not many user
rated or tagged an itembull Lack of Diversity (overfitting) when adapting too strongly to the preferences of users they
might see samesimilar recommendationsbull Use the right context users do things which might not be relevant for their user model
eg try out things do stuff for other peoplebull Research challenge right balance between serendipity amp personalizationbull Research challenge right way to use the influence of recommendations on userrsquos behavior
Social Web 2016 Lora Aroyo and Davide Ceolin
Announcementsbull Next deadlinesbull Tuesday March 1st 2359
Assignment 2bull Friday March 4th 1000 Post
questionbull Friday March 4th 1700 Vote
questions
image source httpwwwflickrcomphotosbionicteaching1375254387
Hands-on Teaser
bull Your Facebook Friendsrsquo popularity in a spread sheet
bull Locations of your Facebook Friendsbull Tag Cloud of your wall posts
- Social Web 2016
- Slide 2
- Slide 3
- Slide 4
- Slide 5
- Slide 6
- Slide 7
- Slide 8
- Slide 9
- Slide 10
- Slide 11
- Slide 12
- Slide 13
- Slide 14
- Slide 15
- Slide 16
- Slide 17
- Slide 18
- Slide 19
- Slide 20
- Slide 21
- Slide 22
- Slide 23
- Slide 24
- Slide 25
- Slide 26
- Slide 27
- Slide 28
- Slide 29
- Slide 30
- Slide 31
- Slide 32
- Slide 33
- Slide 34
- Slide 35
- Slide 36
- Slide 37
- Slide 38
- Slide 39
- Slide 40
- Slide 41
- Slide 42
- Slide 43
- Slide 44
- Slide 45
- Slide 46
- Slide 47
- Slide 48
- Announcements
- Slide 50
-
httpwwwwiredcommagazine201111mf_artsyall1
Social Web 2016 Lora Aroyo and Davide Ceolin
Collaborative Filteringbull Memory-based User-Item matrix ratingspreferences of users =gt
compute similarity between users amp recommend items of similar users
bull Model-based Item-Item matrix similarity (eg based on user ratings) between items =gt recommend items that are similar to the ones the user likes
bull Model-based Clustering cluster users according to their preferences =gt recommend items of users that belong to the same cluster
bull Model-based Bayesian networks P(u likes item B | u likes item A) = how likely is it that a user who likes item A will like item B learn probabilities from user ratingspreferences
bull Others rule-based other data mining techniques
Social Web 2016 Lora Aroyo and Davide Ceolin
bull complete input data is required
bull pre-computation not possible
bull does not scale well bull high quality of
recommendations
bull abstraction (model) of input data
bull pre-computation (partially) possible (model has to be re-built from time to time)
bull scales betterbull abstraction may reduce
recommendation quality
Memory vs Model-based
Social Web 2016 Lora Aroyo and Davide Ceolin
bull collaborative filtering lsquoneighborhoodsrsquo of people with similar interest amp recommending items based on likings in neighborhood
bull limitations next to lsquocold startrsquo and lsquosparsityrsquo the lack of control (over onersquos neighborhood) is also a problem ie cannot add lsquotrustedrsquo people nor exclude lsquostrangersquo ones
bull therefore interest in lsquosocial recommendersrsquo where presence of social connections defines the similarity in interests (eg social tagging CiteULike)bull does a social connection indicate user interest similaritybull how much users interest similarity depends on the strength
of their connectionbull is it feasible to use a social network as a personalized
recommendation[Lin amp Brusilovsky Social Networks and Interest Similarity The Case of CiteULike HTrsquo10]
Social Networks amp Interest Similarity
bull unilaterally connected pairs have more common itemsmetadatatags than non-connected pairsbull highest similarity for direct connections - decreasing with the increase of distance between users in SN bull reciprocal relationship users - significantly larger similarity than users in a unidirectional relationship bull traditional item-level similarity may be less reliable to find similar users in social bookmarking systemsbull peers connected by self-defined social connections could be a useful source for cross-recommendation
Conclusions
Social Web 2016 Lora Aroyo and Davide Ceolin
bull Input characteristics of items amp interests of a user into characteristics of items =gt Recommend items that feature characteristics which meet the userrsquos interests
bull Techniquesbull Data mining methods Cluster items based on their
characteristics =gt Infer usersrsquo interests into clustersbull IR methods Represent items amp users as term vectors
=gt Compute similarity between user profile vector and items
bull Utility-based methods Utility function that gets an item as input the parameters of the utility function are customized via preferences of a user
Content-based Recommendations
Social Web 2016 Lora Aroyo and Davide Ceolin
based on slides from Fabien Abel
Content Features
based on slides from Fabien Abel
User Model
based on slides from Fabien Abel
Recommendations
RecSys Issuesbull Cold-start problem (new user problem) nolittle data available to infer preferences of new
usersbull Changing User Preferences user interests may change over timebull Sparsity problem (new item problem) item descriptions are sparse eg not many user
rated or tagged an itembull Lack of Diversity (overfitting) when adapting too strongly to the preferences of users they
might see samesimilar recommendationsbull Use the right context users do things which might not be relevant for their user model
eg try out things do stuff for other peoplebull Research challenge right balance between serendipity amp personalizationbull Research challenge right way to use the influence of recommendations on userrsquos behavior
Social Web 2016 Lora Aroyo and Davide Ceolin
Announcementsbull Next deadlinesbull Tuesday March 1st 2359
Assignment 2bull Friday March 4th 1000 Post
questionbull Friday March 4th 1700 Vote
questions
image source httpwwwflickrcomphotosbionicteaching1375254387
Hands-on Teaser
bull Your Facebook Friendsrsquo popularity in a spread sheet
bull Locations of your Facebook Friendsbull Tag Cloud of your wall posts
- Social Web 2016
- Slide 2
- Slide 3
- Slide 4
- Slide 5
- Slide 6
- Slide 7
- Slide 8
- Slide 9
- Slide 10
- Slide 11
- Slide 12
- Slide 13
- Slide 14
- Slide 15
- Slide 16
- Slide 17
- Slide 18
- Slide 19
- Slide 20
- Slide 21
- Slide 22
- Slide 23
- Slide 24
- Slide 25
- Slide 26
- Slide 27
- Slide 28
- Slide 29
- Slide 30
- Slide 31
- Slide 32
- Slide 33
- Slide 34
- Slide 35
- Slide 36
- Slide 37
- Slide 38
- Slide 39
- Slide 40
- Slide 41
- Slide 42
- Slide 43
- Slide 44
- Slide 45
- Slide 46
- Slide 47
- Slide 48
- Announcements
- Slide 50
-
Collaborative Filteringbull Memory-based User-Item matrix ratingspreferences of users =gt
compute similarity between users amp recommend items of similar users
bull Model-based Item-Item matrix similarity (eg based on user ratings) between items =gt recommend items that are similar to the ones the user likes
bull Model-based Clustering cluster users according to their preferences =gt recommend items of users that belong to the same cluster
bull Model-based Bayesian networks P(u likes item B | u likes item A) = how likely is it that a user who likes item A will like item B learn probabilities from user ratingspreferences
bull Others rule-based other data mining techniques
Social Web 2016 Lora Aroyo and Davide Ceolin
bull complete input data is required
bull pre-computation not possible
bull does not scale well bull high quality of
recommendations
bull abstraction (model) of input data
bull pre-computation (partially) possible (model has to be re-built from time to time)
bull scales betterbull abstraction may reduce
recommendation quality
Memory vs Model-based
Social Web 2016 Lora Aroyo and Davide Ceolin
bull collaborative filtering lsquoneighborhoodsrsquo of people with similar interest amp recommending items based on likings in neighborhood
bull limitations next to lsquocold startrsquo and lsquosparsityrsquo the lack of control (over onersquos neighborhood) is also a problem ie cannot add lsquotrustedrsquo people nor exclude lsquostrangersquo ones
bull therefore interest in lsquosocial recommendersrsquo where presence of social connections defines the similarity in interests (eg social tagging CiteULike)bull does a social connection indicate user interest similaritybull how much users interest similarity depends on the strength
of their connectionbull is it feasible to use a social network as a personalized
recommendation[Lin amp Brusilovsky Social Networks and Interest Similarity The Case of CiteULike HTrsquo10]
Social Networks amp Interest Similarity
bull unilaterally connected pairs have more common itemsmetadatatags than non-connected pairsbull highest similarity for direct connections - decreasing with the increase of distance between users in SN bull reciprocal relationship users - significantly larger similarity than users in a unidirectional relationship bull traditional item-level similarity may be less reliable to find similar users in social bookmarking systemsbull peers connected by self-defined social connections could be a useful source for cross-recommendation
Conclusions
Social Web 2016 Lora Aroyo and Davide Ceolin
bull Input characteristics of items amp interests of a user into characteristics of items =gt Recommend items that feature characteristics which meet the userrsquos interests
bull Techniquesbull Data mining methods Cluster items based on their
characteristics =gt Infer usersrsquo interests into clustersbull IR methods Represent items amp users as term vectors
=gt Compute similarity between user profile vector and items
bull Utility-based methods Utility function that gets an item as input the parameters of the utility function are customized via preferences of a user
Content-based Recommendations
Social Web 2016 Lora Aroyo and Davide Ceolin
based on slides from Fabien Abel
Content Features
based on slides from Fabien Abel
User Model
based on slides from Fabien Abel
Recommendations
RecSys Issuesbull Cold-start problem (new user problem) nolittle data available to infer preferences of new
usersbull Changing User Preferences user interests may change over timebull Sparsity problem (new item problem) item descriptions are sparse eg not many user
rated or tagged an itembull Lack of Diversity (overfitting) when adapting too strongly to the preferences of users they
might see samesimilar recommendationsbull Use the right context users do things which might not be relevant for their user model
eg try out things do stuff for other peoplebull Research challenge right balance between serendipity amp personalizationbull Research challenge right way to use the influence of recommendations on userrsquos behavior
Social Web 2016 Lora Aroyo and Davide Ceolin
Announcementsbull Next deadlinesbull Tuesday March 1st 2359
Assignment 2bull Friday March 4th 1000 Post
questionbull Friday March 4th 1700 Vote
questions
image source httpwwwflickrcomphotosbionicteaching1375254387
Hands-on Teaser
bull Your Facebook Friendsrsquo popularity in a spread sheet
bull Locations of your Facebook Friendsbull Tag Cloud of your wall posts
- Social Web 2016
- Slide 2
- Slide 3
- Slide 4
- Slide 5
- Slide 6
- Slide 7
- Slide 8
- Slide 9
- Slide 10
- Slide 11
- Slide 12
- Slide 13
- Slide 14
- Slide 15
- Slide 16
- Slide 17
- Slide 18
- Slide 19
- Slide 20
- Slide 21
- Slide 22
- Slide 23
- Slide 24
- Slide 25
- Slide 26
- Slide 27
- Slide 28
- Slide 29
- Slide 30
- Slide 31
- Slide 32
- Slide 33
- Slide 34
- Slide 35
- Slide 36
- Slide 37
- Slide 38
- Slide 39
- Slide 40
- Slide 41
- Slide 42
- Slide 43
- Slide 44
- Slide 45
- Slide 46
- Slide 47
- Slide 48
- Announcements
- Slide 50
-
bull complete input data is required
bull pre-computation not possible
bull does not scale well bull high quality of
recommendations
bull abstraction (model) of input data
bull pre-computation (partially) possible (model has to be re-built from time to time)
bull scales betterbull abstraction may reduce
recommendation quality
Memory vs Model-based
Social Web 2016 Lora Aroyo and Davide Ceolin
bull collaborative filtering lsquoneighborhoodsrsquo of people with similar interest amp recommending items based on likings in neighborhood
bull limitations next to lsquocold startrsquo and lsquosparsityrsquo the lack of control (over onersquos neighborhood) is also a problem ie cannot add lsquotrustedrsquo people nor exclude lsquostrangersquo ones
bull therefore interest in lsquosocial recommendersrsquo where presence of social connections defines the similarity in interests (eg social tagging CiteULike)bull does a social connection indicate user interest similaritybull how much users interest similarity depends on the strength
of their connectionbull is it feasible to use a social network as a personalized
recommendation[Lin amp Brusilovsky Social Networks and Interest Similarity The Case of CiteULike HTrsquo10]
Social Networks amp Interest Similarity
bull unilaterally connected pairs have more common itemsmetadatatags than non-connected pairsbull highest similarity for direct connections - decreasing with the increase of distance between users in SN bull reciprocal relationship users - significantly larger similarity than users in a unidirectional relationship bull traditional item-level similarity may be less reliable to find similar users in social bookmarking systemsbull peers connected by self-defined social connections could be a useful source for cross-recommendation
Conclusions
Social Web 2016 Lora Aroyo and Davide Ceolin
bull Input characteristics of items amp interests of a user into characteristics of items =gt Recommend items that feature characteristics which meet the userrsquos interests
bull Techniquesbull Data mining methods Cluster items based on their
characteristics =gt Infer usersrsquo interests into clustersbull IR methods Represent items amp users as term vectors
=gt Compute similarity between user profile vector and items
bull Utility-based methods Utility function that gets an item as input the parameters of the utility function are customized via preferences of a user
Content-based Recommendations
Social Web 2016 Lora Aroyo and Davide Ceolin
based on slides from Fabien Abel
Content Features
based on slides from Fabien Abel
User Model
based on slides from Fabien Abel
Recommendations
RecSys Issuesbull Cold-start problem (new user problem) nolittle data available to infer preferences of new
usersbull Changing User Preferences user interests may change over timebull Sparsity problem (new item problem) item descriptions are sparse eg not many user
rated or tagged an itembull Lack of Diversity (overfitting) when adapting too strongly to the preferences of users they
might see samesimilar recommendationsbull Use the right context users do things which might not be relevant for their user model
eg try out things do stuff for other peoplebull Research challenge right balance between serendipity amp personalizationbull Research challenge right way to use the influence of recommendations on userrsquos behavior
Social Web 2016 Lora Aroyo and Davide Ceolin
Announcementsbull Next deadlinesbull Tuesday March 1st 2359
Assignment 2bull Friday March 4th 1000 Post
questionbull Friday March 4th 1700 Vote
questions
image source httpwwwflickrcomphotosbionicteaching1375254387
Hands-on Teaser
bull Your Facebook Friendsrsquo popularity in a spread sheet
bull Locations of your Facebook Friendsbull Tag Cloud of your wall posts
- Social Web 2016
- Slide 2
- Slide 3
- Slide 4
- Slide 5
- Slide 6
- Slide 7
- Slide 8
- Slide 9
- Slide 10
- Slide 11
- Slide 12
- Slide 13
- Slide 14
- Slide 15
- Slide 16
- Slide 17
- Slide 18
- Slide 19
- Slide 20
- Slide 21
- Slide 22
- Slide 23
- Slide 24
- Slide 25
- Slide 26
- Slide 27
- Slide 28
- Slide 29
- Slide 30
- Slide 31
- Slide 32
- Slide 33
- Slide 34
- Slide 35
- Slide 36
- Slide 37
- Slide 38
- Slide 39
- Slide 40
- Slide 41
- Slide 42
- Slide 43
- Slide 44
- Slide 45
- Slide 46
- Slide 47
- Slide 48
- Announcements
- Slide 50
-
bull collaborative filtering lsquoneighborhoodsrsquo of people with similar interest amp recommending items based on likings in neighborhood
bull limitations next to lsquocold startrsquo and lsquosparsityrsquo the lack of control (over onersquos neighborhood) is also a problem ie cannot add lsquotrustedrsquo people nor exclude lsquostrangersquo ones
bull therefore interest in lsquosocial recommendersrsquo where presence of social connections defines the similarity in interests (eg social tagging CiteULike)bull does a social connection indicate user interest similaritybull how much users interest similarity depends on the strength
of their connectionbull is it feasible to use a social network as a personalized
recommendation[Lin amp Brusilovsky Social Networks and Interest Similarity The Case of CiteULike HTrsquo10]
Social Networks amp Interest Similarity
bull unilaterally connected pairs have more common itemsmetadatatags than non-connected pairsbull highest similarity for direct connections - decreasing with the increase of distance between users in SN bull reciprocal relationship users - significantly larger similarity than users in a unidirectional relationship bull traditional item-level similarity may be less reliable to find similar users in social bookmarking systemsbull peers connected by self-defined social connections could be a useful source for cross-recommendation
Conclusions
Social Web 2016 Lora Aroyo and Davide Ceolin
bull Input characteristics of items amp interests of a user into characteristics of items =gt Recommend items that feature characteristics which meet the userrsquos interests
bull Techniquesbull Data mining methods Cluster items based on their
characteristics =gt Infer usersrsquo interests into clustersbull IR methods Represent items amp users as term vectors
=gt Compute similarity between user profile vector and items
bull Utility-based methods Utility function that gets an item as input the parameters of the utility function are customized via preferences of a user
Content-based Recommendations
Social Web 2016 Lora Aroyo and Davide Ceolin
based on slides from Fabien Abel
Content Features
based on slides from Fabien Abel
User Model
based on slides from Fabien Abel
Recommendations
RecSys Issuesbull Cold-start problem (new user problem) nolittle data available to infer preferences of new
usersbull Changing User Preferences user interests may change over timebull Sparsity problem (new item problem) item descriptions are sparse eg not many user
rated or tagged an itembull Lack of Diversity (overfitting) when adapting too strongly to the preferences of users they
might see samesimilar recommendationsbull Use the right context users do things which might not be relevant for their user model
eg try out things do stuff for other peoplebull Research challenge right balance between serendipity amp personalizationbull Research challenge right way to use the influence of recommendations on userrsquos behavior
Social Web 2016 Lora Aroyo and Davide Ceolin
Announcementsbull Next deadlinesbull Tuesday March 1st 2359
Assignment 2bull Friday March 4th 1000 Post
questionbull Friday March 4th 1700 Vote
questions
image source httpwwwflickrcomphotosbionicteaching1375254387
Hands-on Teaser
bull Your Facebook Friendsrsquo popularity in a spread sheet
bull Locations of your Facebook Friendsbull Tag Cloud of your wall posts
- Social Web 2016
- Slide 2
- Slide 3
- Slide 4
- Slide 5
- Slide 6
- Slide 7
- Slide 8
- Slide 9
- Slide 10
- Slide 11
- Slide 12
- Slide 13
- Slide 14
- Slide 15
- Slide 16
- Slide 17
- Slide 18
- Slide 19
- Slide 20
- Slide 21
- Slide 22
- Slide 23
- Slide 24
- Slide 25
- Slide 26
- Slide 27
- Slide 28
- Slide 29
- Slide 30
- Slide 31
- Slide 32
- Slide 33
- Slide 34
- Slide 35
- Slide 36
- Slide 37
- Slide 38
- Slide 39
- Slide 40
- Slide 41
- Slide 42
- Slide 43
- Slide 44
- Slide 45
- Slide 46
- Slide 47
- Slide 48
- Announcements
- Slide 50
-
bull unilaterally connected pairs have more common itemsmetadatatags than non-connected pairsbull highest similarity for direct connections - decreasing with the increase of distance between users in SN bull reciprocal relationship users - significantly larger similarity than users in a unidirectional relationship bull traditional item-level similarity may be less reliable to find similar users in social bookmarking systemsbull peers connected by self-defined social connections could be a useful source for cross-recommendation
Conclusions
Social Web 2016 Lora Aroyo and Davide Ceolin
bull Input characteristics of items amp interests of a user into characteristics of items =gt Recommend items that feature characteristics which meet the userrsquos interests
bull Techniquesbull Data mining methods Cluster items based on their
characteristics =gt Infer usersrsquo interests into clustersbull IR methods Represent items amp users as term vectors
=gt Compute similarity between user profile vector and items
bull Utility-based methods Utility function that gets an item as input the parameters of the utility function are customized via preferences of a user
Content-based Recommendations
Social Web 2016 Lora Aroyo and Davide Ceolin
based on slides from Fabien Abel
Content Features
based on slides from Fabien Abel
User Model
based on slides from Fabien Abel
Recommendations
RecSys Issuesbull Cold-start problem (new user problem) nolittle data available to infer preferences of new
usersbull Changing User Preferences user interests may change over timebull Sparsity problem (new item problem) item descriptions are sparse eg not many user
rated or tagged an itembull Lack of Diversity (overfitting) when adapting too strongly to the preferences of users they
might see samesimilar recommendationsbull Use the right context users do things which might not be relevant for their user model
eg try out things do stuff for other peoplebull Research challenge right balance between serendipity amp personalizationbull Research challenge right way to use the influence of recommendations on userrsquos behavior
Social Web 2016 Lora Aroyo and Davide Ceolin
Announcementsbull Next deadlinesbull Tuesday March 1st 2359
Assignment 2bull Friday March 4th 1000 Post
questionbull Friday March 4th 1700 Vote
questions
image source httpwwwflickrcomphotosbionicteaching1375254387
Hands-on Teaser
bull Your Facebook Friendsrsquo popularity in a spread sheet
bull Locations of your Facebook Friendsbull Tag Cloud of your wall posts
- Social Web 2016
- Slide 2
- Slide 3
- Slide 4
- Slide 5
- Slide 6
- Slide 7
- Slide 8
- Slide 9
- Slide 10
- Slide 11
- Slide 12
- Slide 13
- Slide 14
- Slide 15
- Slide 16
- Slide 17
- Slide 18
- Slide 19
- Slide 20
- Slide 21
- Slide 22
- Slide 23
- Slide 24
- Slide 25
- Slide 26
- Slide 27
- Slide 28
- Slide 29
- Slide 30
- Slide 31
- Slide 32
- Slide 33
- Slide 34
- Slide 35
- Slide 36
- Slide 37
- Slide 38
- Slide 39
- Slide 40
- Slide 41
- Slide 42
- Slide 43
- Slide 44
- Slide 45
- Slide 46
- Slide 47
- Slide 48
- Announcements
- Slide 50
-
bull Input characteristics of items amp interests of a user into characteristics of items =gt Recommend items that feature characteristics which meet the userrsquos interests
bull Techniquesbull Data mining methods Cluster items based on their
characteristics =gt Infer usersrsquo interests into clustersbull IR methods Represent items amp users as term vectors
=gt Compute similarity between user profile vector and items
bull Utility-based methods Utility function that gets an item as input the parameters of the utility function are customized via preferences of a user
Content-based Recommendations
Social Web 2016 Lora Aroyo and Davide Ceolin
based on slides from Fabien Abel
Content Features
based on slides from Fabien Abel
User Model
based on slides from Fabien Abel
Recommendations
RecSys Issuesbull Cold-start problem (new user problem) nolittle data available to infer preferences of new
usersbull Changing User Preferences user interests may change over timebull Sparsity problem (new item problem) item descriptions are sparse eg not many user
rated or tagged an itembull Lack of Diversity (overfitting) when adapting too strongly to the preferences of users they
might see samesimilar recommendationsbull Use the right context users do things which might not be relevant for their user model
eg try out things do stuff for other peoplebull Research challenge right balance between serendipity amp personalizationbull Research challenge right way to use the influence of recommendations on userrsquos behavior
Social Web 2016 Lora Aroyo and Davide Ceolin
Announcementsbull Next deadlinesbull Tuesday March 1st 2359
Assignment 2bull Friday March 4th 1000 Post
questionbull Friday March 4th 1700 Vote
questions
image source httpwwwflickrcomphotosbionicteaching1375254387
Hands-on Teaser
bull Your Facebook Friendsrsquo popularity in a spread sheet
bull Locations of your Facebook Friendsbull Tag Cloud of your wall posts
- Social Web 2016
- Slide 2
- Slide 3
- Slide 4
- Slide 5
- Slide 6
- Slide 7
- Slide 8
- Slide 9
- Slide 10
- Slide 11
- Slide 12
- Slide 13
- Slide 14
- Slide 15
- Slide 16
- Slide 17
- Slide 18
- Slide 19
- Slide 20
- Slide 21
- Slide 22
- Slide 23
- Slide 24
- Slide 25
- Slide 26
- Slide 27
- Slide 28
- Slide 29
- Slide 30
- Slide 31
- Slide 32
- Slide 33
- Slide 34
- Slide 35
- Slide 36
- Slide 37
- Slide 38
- Slide 39
- Slide 40
- Slide 41
- Slide 42
- Slide 43
- Slide 44
- Slide 45
- Slide 46
- Slide 47
- Slide 48
- Announcements
- Slide 50
-
based on slides from Fabien Abel
Content Features
based on slides from Fabien Abel
User Model
based on slides from Fabien Abel
Recommendations
RecSys Issuesbull Cold-start problem (new user problem) nolittle data available to infer preferences of new
usersbull Changing User Preferences user interests may change over timebull Sparsity problem (new item problem) item descriptions are sparse eg not many user
rated or tagged an itembull Lack of Diversity (overfitting) when adapting too strongly to the preferences of users they
might see samesimilar recommendationsbull Use the right context users do things which might not be relevant for their user model
eg try out things do stuff for other peoplebull Research challenge right balance between serendipity amp personalizationbull Research challenge right way to use the influence of recommendations on userrsquos behavior
Social Web 2016 Lora Aroyo and Davide Ceolin
Announcementsbull Next deadlinesbull Tuesday March 1st 2359
Assignment 2bull Friday March 4th 1000 Post
questionbull Friday March 4th 1700 Vote
questions
image source httpwwwflickrcomphotosbionicteaching1375254387
Hands-on Teaser
bull Your Facebook Friendsrsquo popularity in a spread sheet
bull Locations of your Facebook Friendsbull Tag Cloud of your wall posts
- Social Web 2016
- Slide 2
- Slide 3
- Slide 4
- Slide 5
- Slide 6
- Slide 7
- Slide 8
- Slide 9
- Slide 10
- Slide 11
- Slide 12
- Slide 13
- Slide 14
- Slide 15
- Slide 16
- Slide 17
- Slide 18
- Slide 19
- Slide 20
- Slide 21
- Slide 22
- Slide 23
- Slide 24
- Slide 25
- Slide 26
- Slide 27
- Slide 28
- Slide 29
- Slide 30
- Slide 31
- Slide 32
- Slide 33
- Slide 34
- Slide 35
- Slide 36
- Slide 37
- Slide 38
- Slide 39
- Slide 40
- Slide 41
- Slide 42
- Slide 43
- Slide 44
- Slide 45
- Slide 46
- Slide 47
- Slide 48
- Announcements
- Slide 50
-
based on slides from Fabien Abel
User Model
based on slides from Fabien Abel
Recommendations
RecSys Issuesbull Cold-start problem (new user problem) nolittle data available to infer preferences of new
usersbull Changing User Preferences user interests may change over timebull Sparsity problem (new item problem) item descriptions are sparse eg not many user
rated or tagged an itembull Lack of Diversity (overfitting) when adapting too strongly to the preferences of users they
might see samesimilar recommendationsbull Use the right context users do things which might not be relevant for their user model
eg try out things do stuff for other peoplebull Research challenge right balance between serendipity amp personalizationbull Research challenge right way to use the influence of recommendations on userrsquos behavior
Social Web 2016 Lora Aroyo and Davide Ceolin
Announcementsbull Next deadlinesbull Tuesday March 1st 2359
Assignment 2bull Friday March 4th 1000 Post
questionbull Friday March 4th 1700 Vote
questions
image source httpwwwflickrcomphotosbionicteaching1375254387
Hands-on Teaser
bull Your Facebook Friendsrsquo popularity in a spread sheet
bull Locations of your Facebook Friendsbull Tag Cloud of your wall posts
- Social Web 2016
- Slide 2
- Slide 3
- Slide 4
- Slide 5
- Slide 6
- Slide 7
- Slide 8
- Slide 9
- Slide 10
- Slide 11
- Slide 12
- Slide 13
- Slide 14
- Slide 15
- Slide 16
- Slide 17
- Slide 18
- Slide 19
- Slide 20
- Slide 21
- Slide 22
- Slide 23
- Slide 24
- Slide 25
- Slide 26
- Slide 27
- Slide 28
- Slide 29
- Slide 30
- Slide 31
- Slide 32
- Slide 33
- Slide 34
- Slide 35
- Slide 36
- Slide 37
- Slide 38
- Slide 39
- Slide 40
- Slide 41
- Slide 42
- Slide 43
- Slide 44
- Slide 45
- Slide 46
- Slide 47
- Slide 48
- Announcements
- Slide 50
-
based on slides from Fabien Abel
Recommendations
RecSys Issuesbull Cold-start problem (new user problem) nolittle data available to infer preferences of new
usersbull Changing User Preferences user interests may change over timebull Sparsity problem (new item problem) item descriptions are sparse eg not many user
rated or tagged an itembull Lack of Diversity (overfitting) when adapting too strongly to the preferences of users they
might see samesimilar recommendationsbull Use the right context users do things which might not be relevant for their user model
eg try out things do stuff for other peoplebull Research challenge right balance between serendipity amp personalizationbull Research challenge right way to use the influence of recommendations on userrsquos behavior
Social Web 2016 Lora Aroyo and Davide Ceolin
Announcementsbull Next deadlinesbull Tuesday March 1st 2359
Assignment 2bull Friday March 4th 1000 Post
questionbull Friday March 4th 1700 Vote
questions
image source httpwwwflickrcomphotosbionicteaching1375254387
Hands-on Teaser
bull Your Facebook Friendsrsquo popularity in a spread sheet
bull Locations of your Facebook Friendsbull Tag Cloud of your wall posts
- Social Web 2016
- Slide 2
- Slide 3
- Slide 4
- Slide 5
- Slide 6
- Slide 7
- Slide 8
- Slide 9
- Slide 10
- Slide 11
- Slide 12
- Slide 13
- Slide 14
- Slide 15
- Slide 16
- Slide 17
- Slide 18
- Slide 19
- Slide 20
- Slide 21
- Slide 22
- Slide 23
- Slide 24
- Slide 25
- Slide 26
- Slide 27
- Slide 28
- Slide 29
- Slide 30
- Slide 31
- Slide 32
- Slide 33
- Slide 34
- Slide 35
- Slide 36
- Slide 37
- Slide 38
- Slide 39
- Slide 40
- Slide 41
- Slide 42
- Slide 43
- Slide 44
- Slide 45
- Slide 46
- Slide 47
- Slide 48
- Announcements
- Slide 50
-
RecSys Issuesbull Cold-start problem (new user problem) nolittle data available to infer preferences of new
usersbull Changing User Preferences user interests may change over timebull Sparsity problem (new item problem) item descriptions are sparse eg not many user
rated or tagged an itembull Lack of Diversity (overfitting) when adapting too strongly to the preferences of users they
might see samesimilar recommendationsbull Use the right context users do things which might not be relevant for their user model
eg try out things do stuff for other peoplebull Research challenge right balance between serendipity amp personalizationbull Research challenge right way to use the influence of recommendations on userrsquos behavior
Social Web 2016 Lora Aroyo and Davide Ceolin
Announcementsbull Next deadlinesbull Tuesday March 1st 2359
Assignment 2bull Friday March 4th 1000 Post
questionbull Friday March 4th 1700 Vote
questions
image source httpwwwflickrcomphotosbionicteaching1375254387
Hands-on Teaser
bull Your Facebook Friendsrsquo popularity in a spread sheet
bull Locations of your Facebook Friendsbull Tag Cloud of your wall posts
- Social Web 2016
- Slide 2
- Slide 3
- Slide 4
- Slide 5
- Slide 6
- Slide 7
- Slide 8
- Slide 9
- Slide 10
- Slide 11
- Slide 12
- Slide 13
- Slide 14
- Slide 15
- Slide 16
- Slide 17
- Slide 18
- Slide 19
- Slide 20
- Slide 21
- Slide 22
- Slide 23
- Slide 24
- Slide 25
- Slide 26
- Slide 27
- Slide 28
- Slide 29
- Slide 30
- Slide 31
- Slide 32
- Slide 33
- Slide 34
- Slide 35
- Slide 36
- Slide 37
- Slide 38
- Slide 39
- Slide 40
- Slide 41
- Slide 42
- Slide 43
- Slide 44
- Slide 45
- Slide 46
- Slide 47
- Slide 48
- Announcements
- Slide 50
-
Announcementsbull Next deadlinesbull Tuesday March 1st 2359
Assignment 2bull Friday March 4th 1000 Post
questionbull Friday March 4th 1700 Vote
questions
image source httpwwwflickrcomphotosbionicteaching1375254387
Hands-on Teaser
bull Your Facebook Friendsrsquo popularity in a spread sheet
bull Locations of your Facebook Friendsbull Tag Cloud of your wall posts
- Social Web 2016
- Slide 2
- Slide 3
- Slide 4
- Slide 5
- Slide 6
- Slide 7
- Slide 8
- Slide 9
- Slide 10
- Slide 11
- Slide 12
- Slide 13
- Slide 14
- Slide 15
- Slide 16
- Slide 17
- Slide 18
- Slide 19
- Slide 20
- Slide 21
- Slide 22
- Slide 23
- Slide 24
- Slide 25
- Slide 26
- Slide 27
- Slide 28
- Slide 29
- Slide 30
- Slide 31
- Slide 32
- Slide 33
- Slide 34
- Slide 35
- Slide 36
- Slide 37
- Slide 38
- Slide 39
- Slide 40
- Slide 41
- Slide 42
- Slide 43
- Slide 44
- Slide 45
- Slide 46
- Slide 47
- Slide 48
- Announcements
- Slide 50
-
image source httpwwwflickrcomphotosbionicteaching1375254387
Hands-on Teaser
bull Your Facebook Friendsrsquo popularity in a spread sheet
bull Locations of your Facebook Friendsbull Tag Cloud of your wall posts
- Social Web 2016
- Slide 2
- Slide 3
- Slide 4
- Slide 5
- Slide 6
- Slide 7
- Slide 8
- Slide 9
- Slide 10
- Slide 11
- Slide 12
- Slide 13
- Slide 14
- Slide 15
- Slide 16
- Slide 17
- Slide 18
- Slide 19
- Slide 20
- Slide 21
- Slide 22
- Slide 23
- Slide 24
- Slide 25
- Slide 26
- Slide 27
- Slide 28
- Slide 29
- Slide 30
- Slide 31
- Slide 32
- Slide 33
- Slide 34
- Slide 35
- Slide 36
- Slide 37
- Slide 38
- Slide 39
- Slide 40
- Slide 41
- Slide 42
- Slide 43
- Slide 44
- Slide 45
- Slide 46
- Slide 47
- Slide 48
- Announcements
- Slide 50
-