Semantic place localization from narratives

Place narratives provide a rich resource of learning how humans localize places. Place localization can be done in various ways, relative to other spatial referents, and relative to agents and their activities in which these referents may be involved. How can we describe places based on their spatial and semantic relationships to objects, qualities, and activities? How can these relations help us improve automated localization of places implicit in textual descriptions? In this paper, we motivate research on extraction of semantic place localization statements from text corpora which can be used for improving document retrieval and for reconstructing locations. The idea is to combine Semantic Web reasoning with existing geographic information retrieval (GIR) and structural text extraction for this purpose. GIR and Semantic Web technology have matured during the last years, but still largely exist in parallel. Current localization approaches have been focusing on the extraction of unstructured word lists from texts, including toponyms and geographic features, not on human place descriptions on a sentence level.