2. The Big Picture
1
2006-2010: Ph.D.: ‘Disambiguating Identity Web References using Social
Data’. The University of Sheffield
2010-2012: Research Associate at the Knowledge Media Institute, The Open
University
Ph.D. Research Associate Future Work
2006-2010 2010-2012
Digital Identity
Digital Identity
Identity Diffusion
Lifecycles
User Behaviour
Time
Dr Matthew Rowe - Existing Research and Future Research Agenda
3. Digital Identity
Digital Identity
Digital Identity
Identity Diffusion
Lifecycles
User Behaviour
2
Personal information is spread across the Web: (a) identity theft, (b) lateral surveillance
Identity theft costs the UK government £1.2 billion per annum (Get Safe Online, 2010)
Manually tracking web citations is time-consuming and repetitive
57% of web users perform ‘vanity’ searches (Pew Internet Report, 2010)
How can identity web references be disambiguated automatically?
Seed data leveraged from Social Web Systems
Information extracted from candidate citations and semantic model built
Devised three disambiguation methods that combine data mining with semantics
Dr Matthew Rowe - Existing Research and Future Research Agenda
4. Digital Identity
Digital Identity
Digital Identity
Identity Diffusion
Lifecycles
User Behaviour
3
Seed Data generation:
Large overlap between offline social networks and online social networks
Exporting semantic social graphs from disparate social web systems (Twitter, Facebook)
Machine-readable user profile and social network information
Interlinking social graphs from disparate social web systems
Disambiguation methods
Rule-based:infer relations between social data and web resources
Graph-based: random walks over a graph space and clustering
Semi-supervised machine learning: classify web citations and learn from classifications
Findings:
Socialdata provides necessary seed data to disambiguate web citations
Achieve best performance using semi-supervised methods, outperforming several baselines
(unsupervised methods)
Rowe and Ciravegna. Disambiguating Identity Web References using Web 2.0 Data and
Semantics. Journal of Web Semantics. 2010
http://www.matthew-rowe.com/?q=thesis
Dr Matthew Rowe - Existing Research and Future Research Agenda
5. Digital Identity
User Behaviour
Digital Identity
Identity Diffusion
Lifecycles
User Behaviour
4
Attention Patterns on Social Web Systems
How is user behaviour associated with heightened attention?
Developed a machine learning approach to:
Identify seed posts
Predict discussion lengths
User Modelling: social network properties, topical focus, community
affinity
Patterns associated with increased attention:
Twittter:
greater broadcast spectrum
Boards.ie: greater community affinity, focussed users
SAP: less community messages, popular users (frequently provide answers)
Rowe et al. Anticipating Discussion Activity on Community Forums. 3rd IEEE International Conference on Social
Computing, Boston, USA. 2011
Rowe et al. Predicting Discussions on the Social Semantic Web. Extended Semantic Web Conference, Heraklion,
Crete. 2011
Wagner et al. What catches your attention? An empirical study of attention patterns in community forums.
International Conference on Weblogs and Social Media, Dublin, Ireland. 2012
Dr Matthew Rowe - Existing Research and Future Research Agenda
6. Digital Identity
User Behaviour
Digital Identity
Identity Diffusion
Lifecycles
User Behaviour
5
Behaviour Analysis in Online Communities
How can the contextual notion of behaviour be captured?
What is the relation between community behaviour and health?
Modelled user behaviour along six dimensions:
Focus Dispersion, Initiation, Contribution, Popularity, Engagement, Content Quality
Modelled behaviour using semantic web technologies:
Behaviour Ontology – capturing contextual notion of behaviour
Inference rules – identifying the role of a given user
Mined roles, and associated behaviour, on a given platform
Correlated the time-series role composition of communities and with health indicators
Found certain roles to be associated with decreases in community health
E.g. Expert Initiators linked to community churn
Roweet al. Community Analysis through Semantic Rules and Role Composition Derivation. Journal of
Web Semantics (in press). 2012
Dr Matthew Rowe - Existing Research and Future Research Agenda
7. Digital Identity
User Behaviour
Digital Identity
Identity Diffusion
Lifecycles
User Behaviour
6
Churn
Churn is the loss of users from a service (telecommunications/social network,
online community)
Goal: predict churners and identify churn patterns
Using social network features (i.e. centrality) provided accurate information for
churn detection
Found:
Differing churn patterns between communities
Central users churn in some communities, while peripheral users churn in others
Currently exploring: Churn diffusion and topological effects
Karnstedtet al. The Effect of User Features on Churn in Social Networks. ACM Web Science
Conference 2011, Koblenz, Germany. 2011
Dr Matthew Rowe - Existing Research and Future Research Agenda
8. The Big Picture - Revisit
7
Ph.D. Research Associate Future Work
2006-2010 2010-2012
Digital Identity
Digital Identity
Identity Diffusion
Lifecycles
User Behaviour
Time
Dr Matthew Rowe - Existing Research and Future Research Agenda
9. Digital Identity
Identity Lifecycles
Digital Identity
Identity Diffusion
Lifecycles
User Behaviour
8
Identity is developed and shaped over time through developmental stages (Eriksson,
1959)
Ego-identity is the ideal that people pursue, while identity is a person’s present state
(Bosma et al., 1994)
How are digital identities shaped online? Do the stages resonate with Eirksson’s
theories?
What development stages do they go through? Is there a common life cycle across
systems?
In role analysis there are common transitions from one role to another
What are the motivations behind digital identity formation and amendments?
Self-efficacy
Self-affirmation
Understanding identity lifecycles leads to:
Better recommendations (followees, products, content)
Tracking of disseminated personal information
Identifying users susceptible to ‘stealing reality’ attacks (Altshuler et al., 2011)
Dr Matthew Rowe - Existing Research and Future Research Agenda
10. Digital Identity
Identity Diffusion
Digital Identity
Identity Diffusion
Lifecycles
User Behaviour
9
Identity Diffusion is the propagation of identity attributes through social systems
I.e. the adoption of defining characteristics from neighbours
What network effects are associated with identity diffusion?
Smallcore of central users found to be influential in protest recruitment (Gonzalez-Bailon et al., 2011)
Core web sites found to influence the spread of memes (Gomez-Rodriguez et al., 2012)
What is the role of passive/active networks on identity formation?
Behaviour adoption is maintained through social reinforcement (Centola, 2010)
Local-level influence (i.e. homophily, inequity, balancing)
Weak-tied individuals in ego-networks influence adoption (Garg et al., 2011)
Inverse correlation between node influence and degree (Katona et al., 2011)
What effects do community actions have on web presence and subscriber churn?
Onlinecommunity churn (Karnstedt et al., 2010), (Zhang et al., 2010), (Kawale et al., 2010)
Recently studied in the context of ego-networks (Quercia et al., 2012), (Kwak et al., 2011)
Understanding and modelling identity diffusion leads to:
Identification
of links between behaviour and churn from online systems
Enable understanding of reductions in web presence (Online marketing, brand promotion)
Dr Matthew Rowe - Existing Research and Future Research Agenda
11. The Big Picture - Recap
10
Ph.D. Research Associate Future Work
2006-2010 2010-2012
Digital Identity
Digital Identity
Identity Diffusion
Lifecycles
User Behaviour
Time
Dr Matthew Rowe - Existing Research and Future Research Agenda
12. 11 Questions?
http://www.matthew-rowe.com
m.c.rowe@open.ac.uk
Dr Matthew Rowe - Existing Research and Future Research Agenda