microsoft research faculty summit 2008. jennifer golbeck assistant professor, college of information...

14
Microsoft Research Faculty Summit 2008

Upload: owen-barrett

Post on 19-Jan-2016

213 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: Microsoft Research Faculty Summit 2008. Jennifer Golbeck Assistant Professor, College of Information Studies University of Maryland, College Park Social

Microsoft Research Faculty Summit 2008

Page 2: Microsoft Research Faculty Summit 2008. Jennifer Golbeck Assistant Professor, College of Information Studies University of Maryland, College Park Social

Jennifer GolbeckAssistant Professor, College of Information StudiesUniversity of Maryland, College Park

Social Networks on the Semantic Web

Page 3: Microsoft Research Faculty Summit 2008. Jennifer Golbeck Assistant Professor, College of Information Studies University of Maryland, College Park Social

The Semantic Web

One big goal of the Semantic Web: Represent knowledge in a standard, machine processable formatAggregate knowledge distributed across the web into a single knowledge modelReason over that knowledge

Page 4: Microsoft Research Faculty Summit 2008. Jennifer Golbeck Assistant Professor, College of Information Studies University of Maryland, College Park Social

Social Networks on the Semantic Web

FOAF (Friend Of A Friend) A simple ontology for representing information about people and who they know

About 20,000,000 social network profiles are available in FOAF formatApproximately 60% of all semantic web data is FOAF data

Page 5: Microsoft Research Faculty Summit 2008. Jennifer Golbeck Assistant Professor, College of Information Studies University of Maryland, College Park Social

Is FOAF a Success Story?

FOAF is a huge part of the Semantic WebIt is presented frequently as a success story of the Semantic WebBUT - it is only a success if it demonstrates that the goals of the Semantic Web are being achieved

Page 6: Microsoft Research Faculty Summit 2008. Jennifer Golbeck Assistant Professor, College of Information Studies University of Maryland, College Park Social

Experiment: Aggregate FOAF

FOAF is automatically generated by many social networking websites

Advogato Buzznet DeadJournal eCademy FilmTrust

Can these networks be connected through people who have profiles on multiple sites?

GreatestJournal InsaneJournal LiveJournal LJ.Rossia.org Minilog.com Tribe

Page 7: Microsoft Research Faculty Summit 2008. Jennifer Golbeck Assistant Professor, College of Information Studies University of Maryland, College Park Social

Semantics of FOAF

Inverse Functional Propertiesfoaf:mbox_sha1sumfoaf:mboxfoaf:homepagefoaf:icqChatIDfoaf:jabberID

Two people who share a common value for one of these properties are inferred to be the SAME person

foaf:msnChatIDfoaf:weblogfoaf:yahooChatIDfoaf:aimChatID

Page 8: Microsoft Research Faculty Summit 2008. Jennifer Golbeck Assistant Professor, College of Information Studies University of Maryland, College Park Social

Properties of FOAF Networks

Range from small to largeSome networks did not provide full member lists, so we had to crawl them

Network Purpose Members StudiedAvg. DegreeAdvogato Business 2,778.00 13.51Buzznet Photos 208,324.00 1.00DeadJournal Blogging 9,801.00 3.74eCademy Business 61,242.00 3.08FilmTrust Social/Entertainment 1,250.00 1.06GreatestJournal Blogging 36,862.00 33.36InsaneJournal Blogging 1,410.00 13.36LiveJournal Blogging 3,563,267.00 8.38Minilog Blogging 119.00 1.63Rossia Blogging 4,180.00 9.65Tribe Social/Entertainment 218,694.00 9.93

Page 9: Microsoft Research Faculty Summit 2008. Jennifer Golbeck Assistant Professor, College of Information Studies University of Maryland, College Park Social

Network Overlap

Page 10: Microsoft Research Faculty Summit 2008. Jennifer Golbeck Assistant Professor, College of Information Studies University of Maryland, College Park Social
Page 11: Microsoft Research Faculty Summit 2008. Jennifer Golbeck Assistant Professor, College of Information Studies University of Maryland, College Park Social

Overall Results

All networks could be connected to at least 4 other networksOver 16,000 accounts could be mergedSmall overall percentage of users, but this is consistent with the fraction of hubs we expect in a network

Page 12: Microsoft Research Faculty Summit 2008. Jennifer Golbeck Assistant Professor, College of Information Studies University of Maryland, College Park Social

Conclusions for the Semantic Web

FOAF utilizes the logic of OWLReasoners can merge profiles of the same person maintained on different networksFOAF is an interesting realization of the Semantic Web goal of maintaining distributed knowledge that can be unified through reasoning.

Page 13: Microsoft Research Faculty Summit 2008. Jennifer Golbeck Assistant Professor, College of Information Studies University of Maryland, College Park Social

Potential Applications

Social Aggregating and Filtering Applications Recommender SystemsEmail FilteringAdvertisingNews SyndicationAll improved with semantics

Trust in Context

Page 14: Microsoft Research Faculty Summit 2008. Jennifer Golbeck Assistant Professor, College of Information Studies University of Maryland, College Park Social

Questions / Comments

Jennifer [email protected]

http://www.cs.umd.edu/~golbeckhttp://trust.mindswap.org