Latent Geospatial Semantics of Social Media

Abstract

Multimodal understanding of shared content is an important success factor for many Web 2.0 applications and platforms. This article addresses the fundamental question of geo-spatial awareness in social media applications. In this context, we introduce an approach for improved characterization of social media by combining text features (e.g., tags as a prominent example of short, unstructured text labels) with spatial knowledge (e.g., geotags, coordinates of images, and videos). Our model-based framework GeoFolk combines these two aspects in order to construct better algorithms for content management, retrieval, and sharing. We demonstrate in systematic studies the benefits of this combination for a broad spectrum of scenarios related to social media: recommender systems, automatic content organization and filtering, and event detection. Furthermore, we establish a simple and technically sound model that can be seen as a reference baseline for future research in the field of geotagged social media.