Geospatial route extraction from texts

Abstract

The need to collect vast amounts of geospatial data is driven by the emergence of geo-enabled Web applications and the suitability of geospatial data in general to organize information. Given that geospatial data collection and aggregation is a resource intensive task typically left to professionals, we, in this work, advocate the use of information extraction (IE) techniques to derive meaningful geospatial data from plain texts. Initially focusing on travel information, the extracted data can be visualized as routes derived from narratives. As a side effect, the processed text is annotated by this route, which can be seen as an improved geocoding effort. Experimentation shows the adequacy and accuracy of the proposed approach by comparing extracted routes to respective map data.