Geographic summaries from crowdsourced data

In this paper, we present a research prototype for creating geographic summaries using the whereabouts of Foursquare users. Exploiting the density of the venue types in a particular region, the system adds a layer over any typical cartography geographic maps service, creating a first glance summary over the venues sampled from the Foursquare knowledge base. Each summary is represented by a convex hull. The shape is automatically computed according to the venue densities enclosed in the area. The summary is then labeled with the most prominent category or categories. The prominence is given by the observed venue category density. The prototype provides two outputs: a light-weight representation structured in GeoJSON, and a semantic description using the Open Annotation Ontology. We evaluate the quality of the summaries using the Sum of Squared Errors (SSE) and the Jaccard distance. The system is available at http://geosummly.eurecom.fr.