Semantic Web in Group Formation Asma Ounnas Learning Societies Lab School of Electronics and Computer Science The University of Southampton, UK ao05r.

Semantic Web in Group Formation

Asma OunnasLearning Societies LabSchool of Electronics and Computer ScienceThe University of Southampton, UKwww.ecs.soton.ac.uk/[email protected]

Group Formation in e-learning

g1

g2

gN

Formation in terms of the constraints

+ Collaboration Goals

as a set of constraints

StudentsInstructor

Groups

Reasoning Web Summer School, Dresden, 07/09/2007

International

female

leader

I want David to be supervise my group

We want Web-based group formation in e-learning

Personalize the allocation of students

Give a degree of freedom for choosing constraints for the formation

Enable the formation of different types of groups

Reasoning Web Summer School, Dresden, 07/09/2007

Potential of the Semantic Web

Used in Personalization of Learning Objects

Used in Social networks

Used in CoPs and Expert Finders

Reasoning Web Summer School, Dresden, 07/09/2007

Constraint-based Group Formation The allocation of participants to groups based on

some constraints

Collaboration task has a set of goals, each is a set of constraints

Each constraint has a value

Maximize the utility of all constraints within all goals and minimize the deviation of the groups satisfaction => Optimal formation

Reasoning Web Summer School, Dresden, 07/09/2007

Assumptions

non-overlapping group formation

all groups have a similar size

All formed groups are stable while the formation is not announced by the instructor

Reasoning Web Summer School, Dresden, 07/09/2007

Hypothesis

Semantic Web technologies

can effectively automate the process of Web-based constrained group formation in the

e-learning domain.

Reasoning Web Summer School, Dresden, 07/09/2007

Group Formation Process

1. Initiating the formation: the instructor sets the constraints and the students’ list;

2. Identifying the members: analysing the students profiles to identify who should be in which group, the descriptions of the students

3. Negotiating the formation: the negotiation of the students’ allocations to groups involves running a specific algorithm that can satisfy the constraints.

=> Modeling + Constraint Satisfaction

Modeling Variables Task-related: experience, education level, knowledge,

skills, abilities (cognitive and physical), grades, interests, preferences of topics and experts.

Relation-oriented: gender, age, culture (race, ethnicity, national origin), social status, personality and behavioral style, social ties, trust.

Context-related: geographical location, availability schedules, and communication tools.

Reasoning Web Summer School, Dresden, 07/09/2007

Modeling For Group Formation

Group Type Task-related Relation-oriented

Teams All variables All variables

Communities Interests, topic preferences, experience, expert preferences

Expertise relationship, trust

Intensional Networks

Skills, abilities, experience None

Social Networks Interests, topic preferences Social ties, trust

Reasoning Web Summer School, Dresden, 07/09/2007

Implementation

Reasoning Web Summer School, Dresden, 07/09/2007

Groups listGroups list

Jena persistent storage (ontology instances)

Students Interface

(Extended foaf-a-matic)

Students Profiles

(SLP ontology)

Group Generator

(inference rules)Group Formation

Algorithms

Instructor InterfaceInstructor

Constraints

Instructor

Students

Semantic Learner Profile Ontology

Based on FOAF Added 14 classes and 34 properties Uses the University of Southampton CS

modules ontology Uses the Trust ontology Domain ontologies can be added

Reasoning Web Summer School, Dresden, 07/09/2007

Semantic Learner Profile Ontology<rdf:RDF xmlns:rdf="http://www.w3.org/1999/02/22-rdf-syntax-ns#" xmlns:rdfs="http://www.w3.org/2000/01/rdf-schema#" xmlns:foaf="http://xmlns.com/foaf/0.1/" xmlns:slp="http://www.ecs.soton.ac.uk/~ao05r/LearnerProfile/slp.owl">

<foaf:Person rdf:nodeID="asma"> <foaf:name>Asma Ounnas</foaf:name>

<foaf:gender>Female</foaf:gender> <slp:teamRole>leader</slp:teamRole>

<slp:firstLanguage>Arabic</slp:FirstLanguage> <slp:takingModule><slp:moduleID>Web Technologies</slp:moduleID></slp:takingModule> <slp:preferredModule>Knowledge Technologies</slp:preferredModule> <slp:preferredTopic>e-learning</slp:preferredTopic> <slp:interest>Semantic Web</slp:interest> <slp:studentOf><foaf:Person><foaf:name>Hugh Davis</foaf:name></foaf:Person></slp:studentOf>

<slp:classmateOf><foaf:Person><foaf:name>Ilaria Liccardi</foaf:name></foaf:Person></slp:classmateOf> </foaf:Person></rdf:RDF>

Reasoning Web Summer School, Dresden, 07/09/2007

Student User Interface

Reasoning Web Summer School, Dresden, 07/09/2007

QuickTime™ and aTIFF (Uncompressed) decompressor

are needed to see this picture.

Instructor User Interface

Group Generator

CSP solvers to get the optimal formation data mining module for incomplete data Deduction rules eg.

IF John is a captain of the football team, THEN John is a leader

IF Sarah has a high grade in discrete mathematics AND Sarah has a high grade in Logic for CS, THEN Sarah will perform well in formal methods

Reasoning Web Summer School, Dresden, 07/09/2007

Evaluation Metrics framework for evaluating Group formation

Quality Using 3 algorithms for group formation with

incomplete data: without rules vs with rules. No claim for:

Proving that any particular set of constraints leads to better results in terms of the groups’ performance.

Claiming that any particular algorithm leads to best grouping.

Feedback

Questions

Reasoning Web Summer School, Dresden, 07/09/2007

Suggestions

Coments