Enhancing Personalized Document Ranking using Social Information

1. RESEARCH OVERVIEW Social networks (like Facebook and MySpace), collaborative bookmarking systems (like Bibsonomy, Delicious, and CiteULike) and Microblog systems like twitter, offer services such as sharing, commenting, tagging, publishing, rating, retweeting and discussing, that make users increasingly active. Hence, users are more and more connected. Given the tremendous amount of information, witch is generated by these platforms, there is a need to an Information Retrieval (IR) system to automatically answer user’s queries. However, IR system, in this case should take into account additional criteria, such as user’s social networks, user’s interests, user’s preferences, etc. In other words, the classical IR systems should be personalized. In personalized information retrieval, the search process considers a user’s model that covers user’s interest, behavior and history. Commonly, users models are build trough user’s query logs [10], user’s posts (such as tweets, blogs and comments) [15], user’s tags and bookmarking [1, 12, 16]. Consequently, a user is represented by a profile. The user profile is then used in IR system in two main scenarios, “query expansion” [3, 4, 7], or document “re-ranking”[6, 8, 9,