Similarity Search on Spatio-Textual Point Sets

User-generated content on the Web increasingly has a geospatial dimension, opening new opportunities and challenges in location-based services and location-based social networks for mining and analyzing user behaviors and patterns. The applications of such analysis range from recommendation systems to geo-marketing. Motivated by these needs, querying and analyzing spatio-textual data has received a lot of attention over the last years. In this paper, we address the problem of matching point sets based on the spatio-textual objects they contain. This is highly relevant for users associated with geolocated photos and tweets. We formally define this problem as a Spatio-Textual Point-Set Join query, and we introduce its top-k variant. For the efficient treatment of such queries, we extend state-of-the-art algorithms for spatio-textual joins of individual points to the case of point sets. Finally, we extensively evaluate the proposed methods using large scale, real-world datasets from Flickr and Twitter.