kiem verbindt software architectuur - thedatatales.org · text mining service application service...

Kiem Verbindt Software Architectuur11 Nov 2018 17:41:04

Purpose

Kiem Verbindt Software Architectuur 2 / 21

Views

MappingNo viewpoint

DocumentationThis view describes how based upon events, a User journey map can be constructedto be interpreted by a User journey manager.

Our story starts with the eventlog. How this log came to be is not the concern of thisarchitecture. The log contains different events and is possibly enhanced torepresent STEM aspects.

S: SentimentT: TopicE: EmotionM: MotivationOn top of that extra events can be deduced from text.

The STEM enhanced eventlog in turn is used to build a model that can be viewed.

ElementsElement TypeContext data Data ObjectEvent Data ObjectEvent data Data ObjectProcess miner Application ComponentRaw user event log Data ObjectSTEM enhanced user eventlog Data ObjectSTEM eventlog classifier Application ServiceText data Data ObjectUser journey manager Business ActorUser journey map Business ObjectUser story miner Application ServiceUser story model Data Object


ParameterizationNo viewpoint


DocumentationTo enable Mapping and Prediction the STEM eventlog classifier needs classificationmodels. These models describe how to classify a text for one of the aspects.

The models are built by using text mining. For this to succeed, annotated text isneeded. This can be created manually by adding annotations to domain relevanttext quotations.

ElementsElement TypeAccount data Business ObjectContext data Data ObjectEvent Data ObjectEvent data Data ObjectMail conversation Business ObjectOpen data Business ObjectRaw user event log Data ObjectSTEM annotated text Data ObjectSTEM Annotation Business ServiceSTEM classification models Data ObjectSTEM event classifiers Application ComponentSTEM eventlog classifier Application ServiceText data Data ObjectText mining algorithm Application ComponentText mining service Application ServiceUser web trail Business Object


PredictionNo viewpoint

DocumentationPrediction is used to inform agents in the organization real time about specific userstory cases that are not going as intended or to even suggest ways to improve theremainder of the case.

For a large part prediction builds on functionality also used for Mapping the userjourney. You can see this on the left. The same User story model is created but fromthere something else follows.

The User story model itself is used to build rules that enable to predict userbehavior based on past behavior. This can be done directly based on the User storymodel which results in a User prediction model. It is also possible to enhance theUser story model manually so that a distinction between different persona is made.This results in a Persona prediction model.

Both models can be used to predict user behavior based on already recordedbehavior in the form of a user story case log. Based on these predictionsintervention alerts and intervention suggestions can be created.

ElementsElement TypeAction Business ProcessContext data Data ObjectEvent Data ObjectEvent data Data ObjectIntervention alert Business ObjectIntervention suggestion Business ObjectPersona coupling Application ServicePersona enriched User story model Data ObjectPersona Prediction Classifier Application ServicePersona prediction model Data Object


Element TypeProcess miner Application ComponentRaw user event log Data ObjectRaw user story case log Data ObjectRule builder Application ComponentSTEM annotated text Data ObjectSTEM classification models Data ObjectSTEM enhanced user eventlog Data ObjectSTEM enhanced user story case log Data ObjectSTEM event classifiers Application ComponentSTEM eventlog classifier Application ServiceSTEM user case log classifier Application ServiceText data Data ObjectText mining algorithm Application ComponentText mining service Application ServiceUser advisor Business ActorUser interaction Business ProcessUser prediction Data ObjectUser prediction classifier Application ServiceUser prediction model Data ObjectUser predictor Application ServiceUser relations manager Business ActorUser story miner Application ServiceUser story model Data Object


Account dataType Business Object

Information about a specific user or account that can be useful to interpret actionsof the user or agents of the organization.

ActionType Business Process

Actions taken by the User advisor.

Intervention alertType Business Object

An intervention alert, alerts a User relations manager so that he can intervene.

Intervention suggestionType Business Object

Suggestion to the User advisor on how to act towards the user.

Mail conversationType Business Object

Event log where successive mails are events.

Open dataType Business Object

Event data like tweeds on twitter or other information that might explain behaviorof users like the weather on a specific day and location.

STEM AnnotationType Business Service

Annotating text with STEM specific annotations.

S: SentimentT: TopicE: EmotionM: Motivation

User advisorType Business Actor

Business Layer


User interactionType Business Process

Users interacting through touch points creates events.

User journey managerType Business Actor

Manages the user journey for the users

User journey mapType Business Object

Visualization of the User story model. Of course multiple visualizations of the samemodel are possible because of the multi dimensionality of the User story model.

User relations managerType Business Actor

User web trailType Business Object

Event log that shows relevant click events from the user.


Context dataType Data Object

Attributes data that is relevant to the event or to a whole case but is not directlyrelated to the event itself.

EventType Data Object

Refers to one activity instance related to one process instance (case) and onetimestamp; event refers to a case, an activity instance, and a point in time; eventshave attributes; events have a name (the classifier of the event), which default isthe activity it refers to, but that is not mandatory;

Event dataType Data Object

Attributes structured data directly related to the event

Persona couplingType Application Service

Using the user story model, specific paths are associated with specific persona.

Persona enriched User story modelType Data Object

A User Story Model where certain paths are annotated with persona information.

Persona Prediction ClassifierType Application Service

A classifier that builds rules that enable to detect what kind of persona is involvedin a specific user story.

Persona prediction modelType Data Object

A set of rules that enable automatic prediction of the persona involved in a specificuser story.

Process minerType Application Component

Process mining solutions like ProM or BupaR or their commercial counterparts.

Application Layer


Raw user event logType Data Object

The raw user event log contains all data relating to events in the user stories thatwas obtained from various sources. Creating the event log is not part of thisarchitecture because this is specific to the organizations providing the log.

Raw user story case logType Data Object

The Raw user story case log contains all data relating to events in a specific userstory that was obtained from various sources. Creating the Raw user story case logis not part of this architecture because this is specific to the organizations providingthe log.

Rule builderType Application Component

STEM annotated textType Data Object

To create text classification models annotated text is used.

STEM classification modelsType Data Object

There are different types of classification hat are used to enhance the eventlog withadditional knowledge that can be mined from text:S: SentimentT: TopicE: EmotionM: MotivationOn top of that extra events can be deduced from text.

The STEM event classification models each support specific classifications. Forreadability of the model these are modeled as one.

STEM enhanced user eventlogType Data Object

This eventlog contains the structured information from the Raw user event log andin addition the now structured data obtained from Text data through the use of theSTEM eventlog classifier.

STEM enhanced user story case logType Data Object

This case log contains the structured information from the Raw user story case logand in addition the now structured data obtained from Text data through the use ofthe STEM eventlog classifier.


STEM event classifiersType Application Component


The STEM event classifiers each perform specific classifications. For readability ofthe model these different classifiers are modeled as one.

STEM eventlog classifierType Application Service


The STEM eventlog classifier bundles the functionality of the different underlyingclassifiers.

STEM user case log classifierType Application Service


The STEM user case log classifier bundles the functionality of the differentunderlying classifiers.

Text dataType Data Object

Text data usually relates directly to a single event and in that way it can be viewedupon as unstructured event data. It can however relate to multiple events and evento events that are not in the raw user eventlog.


Text mining algorithmType Application Component

Text mining as a broad concept is used for different aspects of STEM classification.Different algorithms will be used but are modeled as one to maintain readability ofthe model. Usually supported learning algorithms are used.

Text mining serviceType Application Service

Text mining as a broad concept is used for different aspects of STEM classification.Different services will be used but are modeled as one to maintain readability of themodel.

User predictionType Data Object

Prediction of the most likely next steps. Based on this an advice or an interventionalert can be given.

User prediction classifierType Application Service

A classifier that builds rules to classify the expected next step from the user.

User prediction modelType Data Object

A set of rules that enable automatic prediction of the next step in a specificcustomer story.

User predictorType Application Service

A prediction algorithm that uses the User prediction model and / or the Personaprediction model to predict the most likely next steps. Based on this an advice or anintervention alert can be given.

User story minerType Application Service

Mines eventlogs and creates models / a model that represent the historic userstories.

User story modelType Data Object

The user story model is a model that "shows" the different user stories as a process.


Access relationType Access relation

Text mining serviceSTEM classification models

SourceTarget


Text mining serviceSTEM annotated text

SourceTarget

Realization relationType Realization relation

Text mining algorithmText mining service

SourceTarget


STEM eventlog classifierRaw user event log

SourceTarget


STEM eventlog classifierSTEM enhanced user eventlog

SourceTarget


STEM event classifiersSTEM classification models

SourceTarget


STEM event classifiersSTEM eventlog classifier

SourceTarget


STEM event classifiersSTEM user case log classifier

SourceTarget


Relations


STEM user case log classifierRaw user story case log

SourceTarget


User story minerSTEM enhanced user eventlog

SourceTarget


Process minerUser story miner

SourceTarget


User story minerUser story model

SourceTarget


User prediction classifierUser story model

SourceTarget

Composition relationType Composition relation

EventEvent data

SourceTarget


EventText data

SourceTarget


EventContext data

SourceTarget


User story modelUser journey map

SourceTarget



User journey managerUser journey map

SourceTarget


Rule builderUser prediction classifier

SourceTarget


User prediction classifierUser prediction model

SourceTarget


STEM user case log classifierSTEM enhanced user story case log

SourceTarget


User predictorUser prediction model

SourceTarget


User predictorSTEM enhanced user story case log

SourceTarget


User predictorUser prediction

SourceTarget


User relations managerIntervention alert

SourceTarget


User relations managerIntervention suggestion

SourceTarget



User advisorIntervention suggestion

SourceTarget


User predictionIntervention alert

SourceTarget


User predictionIntervention suggestion

SourceTarget


Context dataOpen data

SourceTarget


Context dataAccount data

SourceTarget


Text dataMail conversation

SourceTarget


Event dataUser web trail

SourceTarget


Text dataAccount data

SourceTarget


Event dataSource


Mail conversationTarget


STEM AnnotationAccount data

SourceTarget


STEM AnnotationMail conversation

SourceTarget


STEM AnnotationUser web trail

SourceTarget


STEM AnnotationSTEM annotated text

SourceTarget


User interactionEvent

SourceTarget


Rule builderPersona Prediction Classifier

SourceTarget


User story minerSTEM enhanced user eventlog

SourceTarget


Persona Prediction ClassifierPersona prediction model

SourceTarget



User predictorPersona prediction model

SourceTarget

Serving relationType Serving relation

User advisorAction

SourceTarget


Persona couplingUser story model

SourceTarget


Persona couplingPersona enriched User story model

SourceTarget


Persona Prediction ClassifierPersona enriched User story model

SourceTarget


Raw user event logEvent

SourceTarget


Raw user story case logEvent

SourceTarget


Event dataOpen data

SourceTarget

Comment on intervention suggestionType Access relation

User advisorEvent data

SourceTarget


Feed backType Association relation

Event dataSTEM annotated text

SourceTarget

Comment on intervention suggestions is used as annotation to improveclassification models. This requires a standardized set of possible replies from theUser advisor.


kiem verbindt software architectuur - thedatatales.org · text mining service application service...

Documents