Semantic Enhanced Community Modelling to Support Knowledge Sharing

1
Semantic - Enhanced Community Modelling to
Support Knowledge Sharing
School of Computing FACULTY OF ENGINEERING

• Kleanthous Styliani
• www.comp.leeds.ac.uk/stellak
• Supervised by Dr. Dimitrova Vania
• User Modelling User Adaptive Systems Group
• Knowledge Representation Reasoning Research
  Group

2
School of Computing FACULTY OF ENGINEERING

• Overview

• Virtual Communities Theoretical Background
• Proposed Framework
• Algorithms
• Study Initial Results
• Community
• Relationship Model
• Centrality
• Individual User Model

Nov 16th 2007 FIT Sankt Augustin
3
Virtual Communities Theoretical
Background Proposed Framework Algorithms Study
Initial Results

• What is a Virtual Community?
• An online community could be a set of users who
  communicate using computer-mediated communication
  and have common interests, shared goals, and
  shared resources. (Preece, 2001)

Loosely Structured
Closely-Knit

• Del.icio.us
• MetaCafe
• MetaFilter

• Educational
• Organisational

Nov 16th 2007 FIT Sankt Augustin
4
Virtual Communities Theoretical
Background Proposed Framework Algorithms Study
Initial Results
Types of Communities

• Closely Knit
• Controlled Membership
• Common Purpose
• Shared Interests
• Sharing information

• Generation of new Knowledge
• High level of dialogue
• Collaboration
• Equal Membership

Nov 16th 2007 FIT Sankt Augustin
5
Virtual Communities Theoretical
Background Proposed Framework Algorithms Study
Initial Results

• The need for Support
• (Fischer and Ostwald, 2001)
• A common misconception is to believe that VC will
  be effective when people and technology are
  present.
• Duplication of resources
• Newcomers integration
• Active members Inactive members
• Ignorance of others knowledge

Nov 16th 2007 FIT Sankt Augustin
6
Virtual Communities Theoretical
Background Proposed Framework Algorithms Study
Initial Results

• Our Approach
• Develop intelligent techniques to automatically
  detect required support tailored to the community
  
• Adaptive Support
• VC as an entity

Nov 16th 2007 FIT Sankt Augustin
7
Virtual Communities Theoretical
Background Proposed Framework Algorithms Study
Initial Results

• Organisational Psychology

Research
Teams
Virtual Community
Relevant Processes
Trust Motivation Shared Mental
Models Transactive Memory Bonding Cognitive
Centrality Adapting Cognitive Consensus Etc
Nov 16th 2007 FIT Sankt Augustin
8
Virtual Communities Theoretical
Background Proposed Framework Algorithms Study
Initial Results
I know many things. But are they interested?
I know all about the central topic here!
Support Needed

Do they know what I know?
Oh! Thats what you meant!
Cognitive Centrality
Yes! I am sure that we have the same understanding
Transactive Memory
I am new here. What they have been doing before?
Cognitive Consensus
Shared Mental Models
Nov 16th 2007 FIT Sankt Augustin
9
Virtual Communities Theoretical
Background Proposed Framework Algorithms Study
Initial Results
Community Model
Community Model Application
Community Model Acquisition
CCs
TM
CCen
SMM
Nov 16th 2007 FIT Sankt Augustin
10
Virtual Communities Theoretical
Background Proposed Framework Algorithms Study
Initial Results
BSCW Example
Nov 16th 2007 FIT Sankt Augustin
11
Virtual Communities Theoretical
Background Proposed Framework Algorithms Study
Initial Results

• Input for Community Model

ENVIRONMENT E
HF Taxonomy of Folders
Folder F

Resource R

RCreatedData RRating, RCreator, RDate, RAssessor, RReader
Based on Dublin Core Metadata element set
RMetadata RDatePublish
Member M

Nov 16th 2007 FIT Sankt Augustin
12
Virtual Communities Theoretical
Background Proposed Framework Algorithms Study
Initial Results

• Community Model

UploadSim
ReadRes
Participation
User Interests
Relationships Model
Cognitive Centrality
Individual User Models
InterestSim
ReadSim
Relationships
Personal Hierarchies
Cognitively Central Members
Community Context
Popular Topics
Peripheral Topics
Nov 16th 2007 FIT Sankt Augustin
13
Virtual Communities Theoretical
Background Proposed Framework Algorithms Study
Initial Results
Modelling Relationships
WordNet
ReadRes Relationship because A read resources
uploaded by B
Nov 16th 2007 FIT Sankt Augustin
14
Virtual Communities Theoretical
Background Proposed Framework Algorithms Study
Initial Results
Modelling Relationships
WordNet
ReadSim UploadSim ReadSim Relationship because
A reads resources similar to those B
reads. UploadSim Relationship because A uploads
resources similar to those B uploads.
Nov 16th 2007 FIT Sankt Augustin
15
Virtual Communities Theoretical
Background Proposed Framework Algorithms Study
Initial Results
Modelling Relationships
WordNet
InterestSim Similarity between two members
interests
Nov 16th 2007 FIT Sankt Augustin
16
Virtual Communities Theoretical
Background Proposed Framework Algorithms Study
Initial Results
Capturing Centrality
Nov 16th 2007 FIT Sankt Augustin
17
Virtual Communities Theoretical
Background Proposed Framework Algorithms Study
Initial Results
The Study

• Data Oct 2005 Dec 2006
• BSCW data anonymised converted into .txt
• Extracted data using Java
• Data stored on a MySQL Database
• Input to algorithms to extract the Community Model

Nov 16th 2007 FIT Sankt Augustin
18

• Initial Results
• Community
• Relationship Model
• Centrality
• Individual User Model

Overview
Nov 16th 2007 FIT Sankt Augustin
19

• Initial Results
• Community
• Relationship Model
• Centrality
• Individual User Model

Activity
01/09/2006 - 31/12/2006
01/06/2006 - 31/08/2006
01/03/2006 - 31/05/2006
01/01/2006 - 28/02/2006
01/10/2005 - 31/12/2005
Nov 16th 2007 FIT Sankt Augustin
20

• Initial Results
• Community
• Relationship Model
• Centrality
• Individual User Model

Uploading
Nov 16th 2007 FIT Sankt Augustin
21

• Initial Results
• Community
• Relationship Model
• Centrality
• Individual User Model

Downloading
01/09/2006 - 31/12/2006
01/06/2006 - 31/08/2006
01/03/2006 - 31/05/2006
01/01/2006 - 28/02/2006
01/10/2005 - 31/12/2005
Nov 16th 2007 FIT Sankt Augustin
22

• Initial Results
• Community
• Relationship Model
• Centrality
• Individual User Model

ReadRes

• Support
• Identify complementary knowledge
• Improve TM
• Encourage Collaboration

Members on the same colour have same number of
ReadRes Red members do not have a ReadRes
relationship
Nov 16th 2007 FIT Sankt Augustin
23

• Initial Results
• Community
• Relationship Model
• Centrality
• Individual User Model

Reading Only

• Have ReadRes with the same members
• Support
• Identify people who are interested in similar or
  same topics
• Make people aware of their similarity
• Encourage Collaboration
• -Building SMM
• -Improve TM

Members on green are only downloading. They all
have a relationship with members in blue
Nov 16th 2007 FIT Sankt Augustin
24

• Initial Results
• Community
• Relationship Model
• Centrality
• Individual User Model

ReadSim

• Support
• Identify relationships that members are not aware
  of
• Who is reading resources similar to those I am
  reading?
• Who is interested in similar resources as I am?
• Improve TM

Nov 16th 2007 FIT Sankt Augustin
25

• Initial Results
• Community
• Relationship Model
• Centrality
• Individual User Model

• Reading resources from the same people but not
  have ReadSim
• Support
• Develop awareness of this similarity
• Encourage contribution
• Improve TM/SMM
• Encourage collaboration
• Facilitate knowledge Sharing

Nov 16th 2007 FIT Sankt Augustin
26

• Initial Results
• Community
• Relationship Model
• Centrality
• Individual User Model

UploadSim

• Very strongly connected
• Support
• Identify people who are not uploading encourage
  them to contribute
• Make people aware of their similarities
• Improve SMM/TM
• Support Collaboration

Nov 16th 2007 FIT Sankt Augustin
27

• Initial Results
• Community
• Relationship Model
• Centrality
• Individual User Model

InterestSim

• Support
• Identify interest similarity complementarities
• Who has interests similar to a given member?
• Motivate contribution
• Encourage collaboration
• Improve SMM/TM

Nov 16th 2007 FIT Sankt Augustin
28

• Initial Results
• Community
• Relationship Model
• Centrality
• Individual User Model

Cognitive Centrality
Support Where important knowledge is located?
Where unique knowledge is located?
Improves TM/SMM Motivation mechanism
Nov 16th 2007 FIT Sankt Augustin
29

• Initial Results
• Community
• Relationship Model
• Centrality
• Individual User Model

• Member 12 uploaded only one resource
• 29.4 of the community read his resource
• Support
• Display similar members, motivate to contribute/
  read
• Use UploadSim to motivate
• Improve TM/SMM

ReadRes Ego Network
UploadSim Ego Network
Nov 16th 2007 FIT Sankt Augustin
30

• Initial Results
• Community
• Relationship Model
• Centrality
• Individual User Model

Newcomer Integration
ReadRes Ego Network of Member 19
Support Identify similar members motivate
this member to contribute Improve
TM/SMM Encourage Collaboration Support Newcomer
Integration
UploadSim InterestSim Ego Network of member 33
Nov 16th 2007 FIT Sankt Augustin
31
Virtual Communities Theoretical
Background Proposed Framework Algorithms Study
Initial Results

• Future Work
• Ontology integration
• What will it be different?
• Community model evaluation
• Using a different VC
• Model community changes over time
• Relationships
• Individual

Nov 16th 2007 FIT Sankt Augustin
32
Virtual Communities Theoretical
Background Proposed Framework Algorithms Study
Initial Results

• Summary
• Problem There is a need to support VC
• Solution Intelligent techniques tailored to the
  whole community can provide the foundations for a
  sustainable VC
• Results TM, SMM, CCen, CCs can be used to
  support VC
• Modelling semantic-enhanced relationships can
  help us to identify what support is needed
• Future plan
• What will be different when the ontology is
  integrated?
• What results can we get if we apply the same
  algorithms in a different VC operating on BSCW
  system?
• What interactions are influencing a VC over time
  and how?

Nov 16th 2007 FIT Sankt Augustin
33
School of Computing FACULTY OF ENGINEERING

• Transactive Memory

The set of knowledge processed by group members,
coupled with and awareness of who knows what.
(Wegner, 1986)

• Transactive Memory in VC
• Be able to
• Know who knows what
• Locate available knowledge
• Beneficial for newcomers

34
School of Computing FACULTY OF ENGINEERING
Shared Mental Models
Members shared and organised understanding and
mental representation of knowledge about key
elements of the teams relevant environment
(Mohammed Dumville, 2001)

• Shared Mental Models in VC
• Form the standards of a community formation
• Development of SMM can improves the effectiveness
  of the group
• Improvement of collaborative knowledge
  exploitation

35
Cognitive Centrality
School of Computing FACULTY OF ENGINEERING

• The greater the degree of overlap between the
  information a member holds and information other
  members hold on average, the greater the degree
  of centrality for that member
  (Kerr Tindale, 2004)

• Cognitive Centrality in VLC
• Central to peripheral What can we do about that?
• Control the community
• Locate unique information

36
Cognitive Consensus
School of Computing FACULTY OF ENGINEERING

• the similarity among group members regarding how
  key issues are defined and conceptualised
  (Mohammed Dumville, 2001)

• Cognitive Consensus in VC
• Same conceptualisation of a concept
• Categorisation/classification of resources

37
School of Computing FACULTY OF ENGINEERING
Capturing Centralitybased on Social Networks
Centrality of Betweenness
Closeness Centrality
Degree Centrality
p3
p3
p3
p2
p2
p4
p4
p4
pk
pk
pk
p5
p5
pn
pn
pn
p5
p2
Relationship
Communication Control
Peripherality
38
School of Computing FACULTY OF ENGINEERING

• This Research

• Research Focus
• Provide holistic personalised support in VC
• Main Assumptions
• Providing adaptation tailored to the community as
  a whole will help the community function better.
• By promoting the building of TM, development of
  SMM, and establishment of CCs and identifying
  CCen inside the community, will improve the
  functioning of this community

39
School of Computing FACULTY OF ENGINEERING

• This Research

Research Questions R1 How to extract a
computational model to represent the functioning
and evolution of the community as a whole, using
semantically enhanced tracking data? R2 Using
that model, how to provide personalised
functionality to support the development of TM,
building of SMM, establishment of CCs and
identification of CCen? R3 How can personalised
support of the above processes affect the
functioning of the community?
40
School of Computing FACULTY OF ENGINEERING

• Existing Systems

41

• Initial Results
• Community
• Relationship Model
• Centrality
• Individual User Model

Newcomer Integration
ReadRes Ego Network of Member 19

• Support
• Use ReadRes to help member integrate.
• Who holds knowledge important to this member?
• Improve TM

42

• Initial Results
• Community
• Relationship Model
• Centrality
• Individual User Model

Integration Problem

• Member 33 uploaded 11 resources
• Never read a resource
• Support
• Help members like 33 to integrate
• Identify similar members motivate this member
  to contribute
• Improve TM/SMM
• Encourage Collaboration
• Support Newcomer Integration

UploadSim InterestSim Ego Network
Ego Network
43

• Initial Results
• Community
• Relationship Model
• Centrality
• Individual User Model

Member 5
Member 2
Member 9

• Only downloading
• Have exactly the same ReadRes relations
• Support
• Encourage collaboration
• Motivate contribution

Nov 16th 2007 FIT Sankt Augustin