exploiting preference queries for searching learning resources fabian abel, eelco herder, philipp...

25
Exploiting Preference Queries for Searching Learning Resources Fabian Abel, Eelco Herder, Philipp Kärger , Daniel Olmedilla, Wolf Siberski L3S Research Center, Hannover, Germany [email protected]

Post on 15-Jan-2016

221 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: Exploiting Preference Queries for Searching Learning Resources Fabian Abel, Eelco Herder, Philipp Kärger, Daniel Olmedilla, Wolf Siberski L3S Research

ExploitingPreference Queries for

Searching Learning Resources

Fabian Abel, Eelco Herder, Philipp Kärger, Daniel Olmedilla,

Wolf Siberski

L3S Research Center, Hannover, Germany

[email protected]

Page 2: Exploiting Preference Queries for Searching Learning Resources Fabian Abel, Eelco Herder, Philipp Kärger, Daniel Olmedilla, Wolf Siberski L3S Research

EC-TEL, September 2007 Philipp Kärger - [email protected]

“ ExploitingPreference Queries forSearching Learning Resources ”

2

Outline

1. What exactly is a preference?

2. A realistic search scenario

3. How preferences help

4. Prototypical implementation

5. Conclusions and future work

Page 3: Exploiting Preference Queries for Searching Learning Resources Fabian Abel, Eelco Herder, Philipp Kärger, Daniel Olmedilla, Wolf Siberski L3S Research

EC-TEL, September 2007 Philipp Kärger - [email protected]

“ ExploitingPreference Queries forSearching Learning Resources ”

3

First, let’s clarify:

What exactly is a

Preference ?

Page 4: Exploiting Preference Queries for Searching Learning Resources Fabian Abel, Eelco Herder, Philipp Kärger, Daniel Olmedilla, Wolf Siberski L3S Research

EC-TEL, September 2007 Philipp Kärger - [email protected]

“ ExploitingPreference Queries forSearching Learning Resources ”

4

– A preference is more than just one preferred value of an attribute

• Simple: “I like green and English”

• Main assumption:– A preference is an order of values

• Better: “I prefer green to red and my last option is brown. I prefer English but German is also fine.”

Page 5: Exploiting Preference Queries for Searching Learning Resources Fabian Abel, Eelco Herder, Philipp Kärger, Daniel Olmedilla, Wolf Siberski L3S Research

EC-TEL, September 2007 Philipp Kärger - [email protected]

“ ExploitingPreference Queries forSearching Learning Resources ”

5

“How can this

help for

technology enhanced learning?”

Page 6: Exploiting Preference Queries for Searching Learning Resources Fabian Abel, Eelco Herder, Philipp Kärger, Daniel Olmedilla, Wolf Siberski L3S Research

EC-TEL, September 2007 Philipp Kärger - [email protected]

“ ExploitingPreference Queries forSearching Learning Resources ”

6

Basic example:

1. “I prefer a cheap course to an expensive one.”

2. “I prefer to have only a few other participants sharing my course instead of an overcrowded course.”

Page 7: Exploiting Preference Queries for Searching Learning Resources Fabian Abel, Eelco Herder, Philipp Kärger, Daniel Olmedilla, Wolf Siberski L3S Research

EC-TEL, September 2007 Philipp Kärger - [email protected]

“ ExploitingPreference Queries forSearching Learning Resources ”

7

part

icip

ants

price

5

10

15

110

20

30

40

20

Page 8: Exploiting Preference Queries for Searching Learning Resources Fabian Abel, Eelco Herder, Philipp Kärger, Daniel Olmedilla, Wolf Siberski L3S Research

EC-TEL, September 2007 Philipp Kärger - [email protected]

“ ExploitingPreference Queries forSearching Learning Resources ”

8

Beyond price and number of participants,learners may have lots of preferences:

• Language an object is presented in

• Where and when does education happen

• By which means (e.g., at a computer or in a reading)

• Who is teaching/authoring

• Type of examination/assessment

• Type of interactivity

• Text or picture-oriented

• …

Page 9: Exploiting Preference Queries for Searching Learning Resources Fabian Abel, Eelco Herder, Philipp Kärger, Daniel Olmedilla, Wolf Siberski L3S Research

EC-TEL, September 2007 Philipp Kärger - [email protected]

“ ExploitingPreference Queries forSearching Learning Resources ”

9

2.A realistic search scenario

Page 10: Exploiting Preference Queries for Searching Learning Resources Fabian Abel, Eelco Herder, Philipp Kärger, Daniel Olmedilla, Wolf Siberski L3S Research

EC-TEL, September 2007 Philipp Kärger - [email protected]

“ ExploitingPreference Queries forSearching Learning Resources ”

10

Page 11: Exploiting Preference Queries for Searching Learning Resources Fabian Abel, Eelco Herder, Philipp Kärger, Daniel Olmedilla, Wolf Siberski L3S Research

EC-TEL, September 2007 Philipp Kärger - [email protected]

“ ExploitingPreference Queries forSearching Learning Resources ”

11

v

• current search approaches:– conjunctive querying: search for an object

bearing all the most preferred attributes– best alternatives act as hard constraints– “return all courses which are on Wednesday

AND take 3 monthsAND with no cost AND …”

in most of the cases no result

Page 12: Exploiting Preference Queries for Searching Learning Resources Fabian Abel, Eelco Herder, Philipp Kärger, Daniel Olmedilla, Wolf Siberski L3S Research

EC-TEL, September 2007 Philipp Kärger - [email protected]

“ ExploitingPreference Queries forSearching Learning Resources ”

12

v

• current search approaches:– disjunctive querying: search for an object

bearing one of all the given properties– e.g., return courses which take 2 months

OR 3 months OR 4 months OR are on Wednesday OR on Monday OR …

will return almost all objects as result

Page 13: Exploiting Preference Queries for Searching Learning Resources Fabian Abel, Eelco Herder, Philipp Kärger, Daniel Olmedilla, Wolf Siberski L3S Research

EC-TEL, September 2007 Philipp Kärger - [email protected]

“ ExploitingPreference Queries forSearching Learning Resources ”

13

preference solution:

we can make use of the given alternatives for each dimension (e.g., if Wednesday is not possible, I go for Monday)

but which courses are optimal according to the preferences?

Page 14: Exploiting Preference Queries for Searching Learning Resources Fabian Abel, Eelco Herder, Philipp Kärger, Daniel Olmedilla, Wolf Siberski L3S Research

EC-TEL, September 2007 Philipp Kärger - [email protected]

“ ExploitingPreference Queries forSearching Learning Resources ”

14

3.How preferences help

finding the desired course

Page 15: Exploiting Preference Queries for Searching Learning Resources Fabian Abel, Eelco Herder, Philipp Kärger, Daniel Olmedilla, Wolf Siberski L3S Research

EC-TEL, September 2007 Philipp Kärger - [email protected]

“ ExploitingPreference Queries forSearching Learning Resources ”

15

The desired courses are Pareto optimal:

A course is optimal if no other course is better (or equal) in all preference dimensions.

example: if a course has the same price but more participants than another, it is not optimal. I.e., the first course is pareto-dominated by the second one

Page 16: Exploiting Preference Queries for Searching Learning Resources Fabian Abel, Eelco Herder, Philipp Kärger, Daniel Olmedilla, Wolf Siberski L3S Research

EC-TEL, September 2007 Philipp Kärger - [email protected]

“ ExploitingPreference Queries forSearching Learning Resources ”

16

No result bears optimal conditions!

Course Weekday Price Distance Location

A Sunday 44 Euro 2 km south

B Friday 44 Euro 2 km south

C Saturday 72 Euro 2 km south

D Saturday no cost 10 km north

E Saturday 72 Euro 10 km north

Page 17: Exploiting Preference Queries for Searching Learning Resources Fabian Abel, Eelco Herder, Philipp Kärger, Daniel Olmedilla, Wolf Siberski L3S Research

EC-TEL, September 2007 Philipp Kärger - [email protected]

“ ExploitingPreference Queries forSearching Learning Resources ”

17

4.

Prototypical Implementation

Page 18: Exploiting Preference Queries for Searching Learning Resources Fabian Abel, Eelco Herder, Philipp Kärger, Daniel Olmedilla, Wolf Siberski L3S Research

EC-TEL, September 2007 Philipp Kärger - [email protected]

“ ExploitingPreference Queries forSearching Learning Resources ”

18

• test data set: 10,000 lectures held at University Hannover

• query language: a novel preference extension of the RDF query language SPARQL

• realized as Web Service integrated in the Personal Reader Framework

Page 19: Exploiting Preference Queries for Searching Learning Resources Fabian Abel, Eelco Herder, Philipp Kärger, Daniel Olmedilla, Wolf Siberski L3S Research

EC-TEL, September 2007 Philipp Kärger - [email protected]

“ ExploitingPreference Queries forSearching Learning Resources ”

19

User Interface

Page 20: Exploiting Preference Queries for Searching Learning Resources Fabian Abel, Eelco Herder, Philipp Kärger, Daniel Olmedilla, Wolf Siberski L3S Research

EC-TEL, September 2007 Philipp Kärger - [email protected]

“ ExploitingPreference Queries forSearching Learning Resources ”

20

Page 21: Exploiting Preference Queries for Searching Learning Resources Fabian Abel, Eelco Herder, Philipp Kärger, Daniel Olmedilla, Wolf Siberski L3S Research

EC-TEL, September 2007 Philipp Kärger - [email protected]

“ ExploitingPreference Queries forSearching Learning Resources ”

22

5.Conclusions and Future Work

Page 22: Exploiting Preference Queries for Searching Learning Resources Fabian Abel, Eelco Herder, Philipp Kärger, Daniel Olmedilla, Wolf Siberski L3S Research

EC-TEL, September 2007 Philipp Kärger - [email protected]

“ ExploitingPreference Queries forSearching Learning Resources ”

23

• Conclusions:

– Classical search mechanisms consider “preferences” as hard constraints

• Problem if no optimal solution exists

– Preference-based queries allow for soft constraining the results

• pruning the non dominated learning resources dramatically decreases the size of the result set

Page 23: Exploiting Preference Queries for Searching Learning Resources Fabian Abel, Eelco Herder, Philipp Kärger, Daniel Olmedilla, Wolf Siberski L3S Research

EC-TEL, September 2007 Philipp Kärger - [email protected]

“ ExploitingPreference Queries forSearching Learning Resources ”

24

• Observation:– Users do not need to specify all preferences

• Only those they want

– Preferences might be automatically extracted• If the student’s schedule is full on Monday then …• If the student’s results are bad for oral exams then …

– Default preferences might be turned on• Cheapest price, with certification, lowest distance, highest

reputation, etc…

Page 24: Exploiting Preference Queries for Searching Learning Resources Fabian Abel, Eelco Herder, Philipp Kärger, Daniel Olmedilla, Wolf Siberski L3S Research

EC-TEL, September 2007 Philipp Kärger - [email protected]

“ ExploitingPreference Queries forSearching Learning Resources ”

25

• Future Work

– extend preference based search withpreference based recommendation

– combine this with established collaborative filtering strategies

• hybrid solution (e.g., to solve cold start problems)

– using preferences in Curriculum planning

Page 25: Exploiting Preference Queries for Searching Learning Resources Fabian Abel, Eelco Herder, Philipp Kärger, Daniel Olmedilla, Wolf Siberski L3S Research

EC-TEL, September 2007 Philipp Kärger - [email protected]

“ ExploitingPreference Queries forSearching Learning Resources ”

27

Thanks for your attention.

Philipp Kärger

L3S Research Center

Hannover, Germany

[email protected]