A Study of the Benet of Leveraging Tagging Behavior to Model Users' Interests in del.icio.us

Social tagging sites such as Flickr, YouTube and del.icio.us are becoming increasingly popular. Users of these sites annotate and endorse content by tagging, and form social ties with other users by including them into their friendship network. The richness of social context raises the users' expectations with respect to the quality of served content, but also presents a unique opportunity for the design of semantically-enriched rec- ommender systems. This paper presents a variety of methods for producing customized hotlists and eval- uates their eectiv eness on del.icio.us datasets. We model a user's interest in terms of the tags he uses to annotate content, and in terms of his explicitly stated and derived social ties, and demostrate how such in- terest can be leveraged to produce hotlists of very high quality. We also discuss possible research directions and outline strategies for the design of a social tagging recommender system.