Découverte d'associations sémantiques pour le Web Sémantique Géospatial - le framework ONTOAST. (Semantic association discovery for the Geospatial Semantic Web - the ONTOAST framework)

It is now commonly accepted that over 70% of web pages contain spatial and temporal referencesthrough the use of place names, addresses, geographical coordinates, dates, etc. However, the temporal and spatialdescriptions are currently unexploited by search engines, when they could be used in the search process for defining thecontext of a query, for query disambiguation, for result classification, etc. Based on this observation, our work focuseson the study of representation and reasoning techniques for spatial and temporal information in the context of the futureGeospatial Semantic Web. The objective of the Geospatial Semantic Web is similar to that of the Semantic Web : attachto spatial and temporal data formal descriptions (metadata) that can be interpreted by humans, but mostly by machines,so that the automated processing of this data by software agents becomes possible and efficient.We propose in this thesis a spatial and temporal reasoner, which is compatible with the standard ontology languageOWL and with the evolution OWL 2. The system, called ONTOAST, is able to exploit both spatial and temporalquantitative data and spatial and temporal qualitative relations in order to infer implicit spatial and temporal qualitativerelations. The goal is to answer questions such as : "What cities are located in the southwest of France ?", "What arethe tourist attractions near my current position ?" . . .This thesis also studies an alternative search paradigm, calledsemantic analysis, which aims the discovery of directand indirect relationships existing between two individuals described using RDF(S) graphs. In order to infer additionalsemantic associations and to increase the accuracy of the analysis, we propose an adaptation of the semantic analysis forOWL 2 ontologies. We also show that new and possibly interesting semantic associations can be discovered, by takinginto account spatio-temporal information which is usually attached to resources. Moreover, we propose to handle spatialand temporal contexts in order to limit the scope of the analysis to a region of space and a period of time. Thesemanticanalysis discovery process uses ONTOAST for reasoning with spatial and temporal information and relations.