Ontology-driven geographical information retrieval

A common problem in information retrieval is that the terms employed by a user to refer to some concept may not be equivalent to the terms employed in a database to refer to the same concept. In a geographical context the user may be interested in the “Hebrides” of Scotland, but a relevant database might refer only to the formal administrative regions, such as “Western Isles Islands”, or to the associated town names such as “Stornaway”. A related problem is that the user may be interested in places in the vicinity of a known place without knowing the names of the nearby places. A means is required therefore to translate between equivalent terminology, find related information and rank the results with regard to their expected user interest (Jones et al 1996; Beard and Sharma 1997; Rodriguez et al 1999). This can be addressed through the use of ontologies and thesauri that encode semantic relationships between concepts and hence facilitate the detection of associations between related terms. Relationships of synonymy and of part-whole are to be found in geographical thesauri such as the Thesaurus of Geographic Names (TGN) (Harpring 1997) and in some gazetteers (Hill et al 1999; Moss et al 1998).