Top-K Frequent Term Queries on Streaming Data

With rapidly increasing user-generated geo-tagged content in social media, location-based queries have received more attention lately. This paper studies the problem of efficiently answering top-K spatial term queries on streaming data where given a term, the goal is to find top K locations where the term is frequent. We propose two variations of reverse spatial term queries on streaming data and an approach for efficiently evaluating them with some bounds on the error. We conduct experiments on a relatively large dataset of Flickr photos, reporting a preliminary evaluation of our approach. We also discuss the limitations of the approaches in the literature for answering reverse spatial term queries and some of the advantages and disadvantages of our proposed approach.