Pain to control Dislikes and distrusted because we don’t know what they listen to, and who they talk to, and what they do with that data.
Publishing web (publishing) – blogging, tweeting, updating their status, etc. Sharing web (sharing) – want others to see what we publish, which groups we’re in, what we like and dislike, our opinion on things, what we’ve been up to, places we’re visitingSocial networks (networking) – becoming part of online communities, of friends, colleagues, or complete strangers And as you know, the social web is increasingly becoming the new renewable energy of RS – it’s cheap, exists in abundance all around us, but largely untamed.Spyware is not the answer, where you develop apps that sit inside computers to monitor and analyze what we browse. Much research went into that direction. Problem is that people lose control on when and what to share and what not to share. Of course they could always switch off/on the spyware but it’s a headache and people worry that they might forget, or don’t trust the tool to behave like it should.
Disconnection of knowledge and social networkWant the SNS to talk to each other so they can give me a better serviceOvertime, the cumulative frequencies of the tags you use canbe represented with a tag-cloud. This gives a visual snapshot of the terms that you use most frequently.When we began this work, the first thing we did was develop a toolFor viewing tag clouds from multiple domains. We noticed thatmany tags represented concepts that could be considered Interests of the users.Hence, the motivation for our work is to exploit this tagging
Facebook is heading the move towards globalising how you learn about what your users like, by allowing them to tell you what they like wherever they are whenever they like. Problem – you don’t know who they will sell this info to, info locked within Facebook
What u like: from browsing, purchasingWhom u know what they like what you might like
The DBpedia knowledge base currently describes more than 3.4 million things, out of which 1.5 million are classified in a consistent Ontology, including 312,000 persons, 413,000 places, 94,000 music albums, 49,000 films, 15,000 video games, 140,000 organizations, 146,000 species and 4,600 diseases. The DBpedia data set features labels and abstracts for these 3.2 million things in up to 92 different languages; 841,000 links to images and 5,081,000 links to external web pages; 9,393,000 external links into other RDF datasets, 565,000 Wikipedia categories, and 75,000 YAGO categories. The DBpedia knowledge base altogether consists of over 1 billion pieces of information (RDF triples) out of which 257 million were extracted from the English edition of Wikipedia and 766 million were extracted from other language editions
Sense here refers to adding meaning to tags, structure, modeling users
Disambiguation based on similarity of term vectors of Dbpedia pages and tag terms based on their frequency.
Interlinking semantics, web2.0, and the real-world HarithAlani Knowledge Media institute, OU APRESW Workshop, Extended Semantic Web Conference, Crete, 2010
Learning about YOU! Sites learn about what you like from your browsing/purchasing history Cold start problem New user New site New product, product range Sparse knowledge Limited to interactions within the site Can’t learn if you are using other sites No to Spyware! Tools that sit inside computers and monitor browsing behaviour and content Much research went into that direction for building RSs Pain to control what they should/shouldn’t access and when 2
New info sources for recommendation systems Micro publishing blogging tweeting updating status messaging Posting 3 publishing
Facebook’s Open Graph Collects “like” information from anywhere about anything! “Based on the structured data you provide via the Open Graph protocol, your pages show up richly across Facebook: in user profiles, within search results and in News Feed.” 5
Personal interests and the social web WHAT YOU LIKE WHOM YOU KNOW 6
Un-Semantic Recommender Systems 8 Collaborative filtering is scalable, relatively cheap, and requires little background knowledge But can semantics help improve recommendation accuracy? Could it be cost effective?
Semantics from Linked Open Data Cloud 2007 millions of objects Billions of triples 9
DBpedia – a Linked Data hub 10 Status: No Relation found
11 Social content Recommender Systems Social networks Semantic web YES … BUT!!
Challenges Tag ambiguity, misspellings, redundancy No semantic structure Distributed and disintegrated personal tag clouds Disconnected social network islands Limited accessibility to data on SNSs 12 publishing networking sharing Live Social Semantics platform aims at solving these problems!
Social+Semantics+RFID: Live Social Semantics Integration of physical presence and online information Semantic user profile generation Interest identification from distributed tagging activities Large-scale, real-world social gatherings Logging of face-to-face contact Social network browsing On-site and post-event support for social networking 13
Making Sense of Folksonomies Semantic User Profiles FOAF DBpedia + Wordnet Identity Integration Tag Integration Delicious Last.fm Flickr Facebook …
Tag Disambiguation Term vector similarity Term vector from tag co-occurrence Term vector for each suggested Dbpedia disambiguation page 23 apple, film, 1980, .. apple, inc, computer, .. apple, iphone, computer, .. apple, tree, fruit, ..
Tags to User Interests Based on 72 POIs verified by users 24
From social to semantics Cleaning up the tag Associating tags with semantics Integrating tagging information Collecting and merging social networks 25
SocioPatterns platform: motivation fundamental knowledge on human contact epidemiological relevance for airborne pathogens communication in mobile scenarios organizational investigation ubiquitous social networking augmented (social) reality 32
Convergence with online social networks 33 leverage social context
Statistics ESWC 2009 attended by over 300 people 187 collected an RFID 139 created accounts on LSS site HyperText 2009 attended by around 150 people 113 collected an RFID 97 registered on LSS site 41
Survey of users who didn’t provide LSS with any SNS accounts 84 registered with no SNS accounts 36 responded to our survey Some used LinkedIn or xing This survey does not include conf attendees who did not participate in LSS 42
Recommendation Services for LSS Recommending talks and sessions If speakers are in your online social network If speakers are in your community of practice network If you have met the speakers during the conference or in past events Recommending people for your online social network If you spent time talking to someone not in your online social network If you met someone who is influential, active If you have strong indirect connections to a person you met Recommending people you should meet If you have strong overlap of interests If your community of practice is very similar If you have an overlapping social network Recommending popular topics/sessions to organisers If a talk/session is heavily attended If a talk/speaker generated much attention 43
Acknowledgement CiroCattuto - ISI Turin Wouter van Den Broeck - ISI Turin Martin Szomszor - CeRC, City University, UK Alain Barrat - CPT Marseille & ISI GianlucaCorrendo – Uni Southampton, UK Organizers of ESWC 2009, HT 2009, and ESWC 2010 Users of LSS! Live Social Semantics references: Szomszor, M., et al. (2010) Semantics, Sensors, and the Social Web: The Live Social Semantics experiments. Extended Semantic Web Conference (ESWC), Crete. Broeck, W., et al. (2010) The Live Social Semantics application: a platform for integrating face-to-face presence with on-line social networking, Workshop on Communication, Collaboration and Social Networking in Pervasive Computing Environments (PerCol), IEEE PerCom, Mannheim. Alani, H., et al. (2009) Live Social Semantics. In: 8th International Semantic Web Conference (ISWC, US. 44
THANKS! please consider participating in LSS http://tagora.ecs.soton.ac.uk 45