Lead


Lay summary
Existing search engines on the Web are weak in supporting spatial queries. While local prepositions, such as "close to", may be used as strings in a query, they are usually evaluated according to the frequency of their occurrence in the indexed documents and not according to their meaning in natural language. When searching for businesses or government agencies in urban centers some engines return resources of real world entities, which are "in the surroundings of" or "close to" a reference place. However, this does not apply to the entire range of resources on the Web and not to rural areas.This project addresses the shortcomings of existing search services by developing a framework for the processing of (possibly vague) spatio-thematic queries, which is generally applicable. The framework uses a declarative formalism in order to represent spatial knowledge. This allows spatial knowledge to be combined with thematic knowledge, which is nowadays usually expressed in declarative Web Ontology Language (OWL). If the represented spatial relations are linked to local prepositions, such as "close to" or "next to", then queries using vague spatial concepts can be processed.The objectives of the proposed project thus are to* develop a theory of vague spatial relations;* explore competing implementation strategies for an application to process (possibly vague) spatio-thematic queries;* evaluate the framework using a prototype application.The first objective is addressed by extending the Region Connection Calculus (RCC), which is a formalism for spatial knowledge representation and inference, and by establishing mappings between frequently used local prepositions and spatial relations. Addressing the second objective involves describing the conditions under which each strategy is successful and identifying the most promising for productive use in applications. Addressing the third objective requires identifying the degree of agreement/disagreement between a set of local prepositions, as perceived by human subjects, and the corresponding spatial relations, as computed by the prototype application, for a set of pairs of spatial features.The successful project will pave the way for a new kind of "intelligent" application that can be used to solve problems requiring both thematic and spatial knowledge. Potential uses include both expert systems and Web 2.0 applications supporting geo-related activities, such as geo-marketing, in industries, such as housing, tourism and media. Potential uses also include applications outside the geospatial domain, for instance, applications that acquire knowledge by extracting information from multimedia content.