Leveraging social information for personalized search

Social information retrieval becomes a very challenging task with the increase use of social networks and the amount of social information they provide continuously in different fields. In this paper, we aim at exploring different kind of social information, namely descriptions (tags) and reactions (clicks) to build user and document profiles for personalization aim. The goal is threefold: (1) propose a social user profile based on community detection considering descriptions, (2) introduce a new notion of social document profile using reactions and (3) propose a personalized ranking model based on social relevance that is computed considering the social document and user profiles. We evaluate our approach on a last.fm dataset using exact matching and approximate matching algorithms. Results show that our approach significantly outperforms the baseline in terms of effectiveness by more than 26% in NDCG@5 for approximate matching and 15% for exact matching. The improvement reaches 43% when only user profile is considered for computing relevance.