An Efficient Index for Top-k Keyword Search on Social Networks

Abstract

Social networks (e.g., Facebook, Twitter) have attracted significant attention recently. Many users have a search requirement to find new or existing friendships with similar interests in social networks. A well-known computing model is keyword search, which provides a user-friendly interface to meet users search demands. However traditional keyword search techniques only consider the textual proximity and ignore the relationship closeness between different users. It is a big challenge to integrate social relationship and textual proximity and it calls for an effective method to support keyword search in social networks. To address these challenges, we present a tree decomposition based hierarchical keyword index structure (TDK-Index) to solve the problem. Our major contributions are: (1)TDK-Index which integrate keyword index and relationship closeness index as a whole; (2)Two-phase TA algorithm which narrows the threshold obviously compared to existing methods and speed up top-k query by a factor of two; and (3)flexible solution which adopts different application circumstances by parameter adjustment. Our experiments provide evidences of the efficiency and scalability of our solution.