Characterization of toponym usages in texts

Toponyms in texts and search queries are often used figuratively and do not directly refer to the locations they reference in their literal sense. Different usage kinds and stylistic devices characterize toponym usages in texts. It is thus crucial for a Geographic Information Retrieval (GIR) system to precisely distinguish these different toponym usages at indexing and at query time in order to best address a given information need and the geospatial footprint of a document. For that purpose, we analyze which of the classic stylistic devices such as allegories, metaphors, or metonymies are used together with toponyms. We use these categories as a foundation for a systematic approach towards the characterization of toponym usages in texts which we believe is necessary to further boost retrieval effectiveness of future GIR systems. A prototype implements this characterization exemplary for texts written in German. We evaluate the effectiveness of our approach against a reference corpus to show the general feasibility. Our approach provides a basis for a wide range of more sophisticated applications such as for example text genre detection.