Our goal: to help the user navigate in a social tagging graph
• Navigation-based query construction
• Online incremental query execution
• Offline and online ranking and clustering
+ @serge + #social
- #marketing
Birds of a tag flock together
Workshop on Search in Social Media 2010
Social tagging data R(user,tag,item)
Affinity between
users, tags and items
Clusters of users, tags and items
CiteULike, delicious, MovieLens Serge Abiteboul, Sihem Amer-Yahia, Alban Galland, Amélie Marian, Pierre Senellart
Item partition
#SIGMOD
#WSDM
Tag partition
User partition
#SSM Item cloud
Tag cloud
User cloud
AmerYahia09Building
Chi07Understanding Harvey09Folksonomic
Joshi09Improving
Li07Towards
Marlow06HT06
Oldenburg08Similarity
SantosNeto07Tracking
#ACMPapers
#SpringerPapers
in
Online Offline
Top-k ranking
Cluster selection
Social graph analysis
Labeled clustering
Challenges and on-going work
• Affinity: domain-dependent definition based on experiments on real dataset
• Top-k: incremental query evaluation based on user interaction
• Labeled clustering: adaptation of subspace clustering to social graph
• Cluster selection: efficient evaluation based on user interaction
• User navigation: affinity-based ranking and zooming
Another query example:
query(@u,(top-10) $X) = Friends(@serge,@u)
Near(G,$X,@u)
Near(G,$X,#CarlaBruni) Near(G,$X,#Sarkozy)
