Geo-Social Keyword Search ⋆

Abstract. In this paper, we propose Geo-Social Keyword (GSK) search, which enables the retrieval of users, points of interest (POIs), or keywords that sat-isfy geographic, social, and/or textual criteria. We first introduce a general GSK framework that covers a wide range of real-world tasks, including advertisement, context-based search, and market analysis. Then, we present three concrete GSK queries: (i) NPRU that returns the top-k users based on their spatial proximity to a given query location, their popularity, and their similarity to an input set of terms; (ii) NSTP that outputs the top-k POIs based on their proximity to a user v, the number of check-ins by friends of v, and their similarity to a set of terms; (iii) FSKR that discovers the top-k keywords based on their frequency in pairs of friends located within a spatial area. For each query, we develop a processing algorithm that utilizes a novel hybrid index. Finally, we evaluate our framework with thorough experiments using real datasets.