This document summarizes a novel application called Live Social Semantics that integrates data from the semantic web, online social networks, and an RFID-based face-to-face contact sensing platform. It tracks face-to-face contacts between conference attendees using RFID badges, builds profiles of attendees' interests by linking their social media data, and allows attendees to view their connections to other attendees. The system was tested at a conference where over 300 attendees participated, and future work is proposed to improve interest profiling and support additional applications using the social interaction data.
1. Live Social Semantics A novel application that integrates data from the semantic web, online social networks, and a real-world face-to-face contact sensing platform. Martin Szomszor University of Southampton
2. Outline History Where Live Social Semantics came from LSS Architecture Tracking Face-to-Face Contacts Integrating and Managing Data Building Profiles of Interest Video Demonstration LSS at ESWC2009 Future Work
4. Sociopatterns.org This projects aims to shed light on patterns in social dynamics and coordinated human activity. We do so by developing and deploying an experimental social interaction sensing platform. This platform consists of portable sensing device and software tools for aggregating, analyzing and visualizing the resulting data. http://www.sciencegallery.com/infectious
7. LSS – Proposed Features Contact Histories “Hey, I remember talking to this person, but I don’t know their name / email / institution” People you might know “Who are the people in my social networks / community of practice who are also attending the conference? What papers are they presenting” Profiles of Interest “I’d like to expose the things that I’m interested in to other participants, including extra-academic data”
13. Active RFID Proximity Detection spatial resolution ~ 1 meter anisotropy - face-to-face temporal resolution ~ 5-20 seconds unobtrusive scalable low cost (~15 Euro per badge – reusable) easily deployable distributed
14. RDF Representation of Contact Data http://tagora.ecs.soton.ac.uk/LiveSocialSemantics/eswc2009/1410 http://tagora.ecs.soton.ac.uk/LiveSocialSemantics/eswc2009/1515 hasPhysicalContact contactWith http://tagora.ecs.soton.ac.uk/LiveSocialSemantics/eswc2009/contact/day3/1410/1515 "2009-06-03"^^<http://www.w3.org/2001/XMLSchema#date> contactDate "00:01:43"^^<http://www.w3.org/2001/XMLSchema#time> contactDuration
15. Architecture COP + Publications RKBExplorer.com Profile Builder dbtune.org Publications data.semanticweb.org dbpedia.org Consumes Tagging Data TAGora Sense Repository Extractor Daemon Delicious Social Tagging Social Networks Web Based Systems Flickr mbid - > dbpediauri tag -> dbpediauri Lastfm Returns Profile of Interests Contacts Facebook Connect API 4store RFID Readers Local Server Social Semantics RDF Cache Aggregator Real World Real World Contact Data RFID Badges
16. How are you connected? Delicious Folksonomies, The Semantic Web, and Movie Recommendation CiroCattuto Martin Szomszor Live Social Semantics Publications www.tagora-project.eu Projects
17. Distinct, Separated Identity Management http://tagora.ecs.soton.ac.uk/LiveSocialSemantics/eswc2009/foaf/1 Martin Szomszor Delicious Tagging and Network RFID Contact Data http://tagora.ecs.soton.ac.uk/delicious/martinszomszor http://tagora.ecs.soton.ac.uk/LiveSocialSemantics/eswc2009/1410 Flickr Tagging and Contacts Conference Publication Data http://tagora.ecs.soton.ac.uk/flickr/7214044@N08@N08 http://data.semanticweb.org/person/martin-szomszor/ Lastfm favourite artists and friends Past Publications, Projects, Communities of Practice http://tagora.ecs.soton.ac.uk/lastfm/count-bassy http://southampton.rkbexplorer.com/id/person-05877 Facebook contacts http://tagora.ecs.soton.ac.uk/facebook/613077109
19. Profile Building 1) Disambiguate Tags cosine similarity between user co-occurrence vector and term frequency vector from concept Choose Sense if above threshold (0.3) or single sense 2) Calculate Interest Weights weight w = fr ∗ ur , where fr is the total frequency of all tags disambiguated to sense r, and ur is a a time decay factor. The factor ur = ⌈days(r)/90⌉ 3) Create Interest List If more than 50 interests are suggested, we rank by weight and suggest the top 50 Users must verify the list before it is published
27. Survey Results After the conference, we emailed the users who did register on our site, but did not enter any social networking accounts. The aim was to understand the reasons why:
28. Future Work Allow individuals to link to their own foaf profiles More SNS sites: Twitter, LinkedIn, etc… Document and Advertise Linked Data Interface Support other applications in exploiting the data Recommend Contacts What features are most predictive of face-to-face contact
29. Building Better Profiles What tags correspond to interests? Locations and topics are useful, but other terms are not TF / IDF Approach It’s not that useful to find out we are all interested in RDF and the Semantic Web Making use of the Category hierarchy If I’m interested in Facebook, Flickr, Last.fm, Delicious, etc, I can extrapolate the interest Online_Social_Networks
30. University of Southampton Acknowledgements CiroCattuto, Wouter Van den Broeck, Alain Barrat HarithAlani, Martin Szomszor, GianlucaCorrendo