Natural-language spatial relations: Metric refinements of topological properties

This research focuses on the communication of spatial relations between a user and a geographic information system (GIS). The novel contribution is the capability of capturing the semantics of spatial predicates within a GIS. This capability is achieved through the development of a formalism that is flexible enough to be calibrated for different application domains, used in different contexts, and cultures. Most current linguistic and cognitive models of spatial relations are based on introspection and lack formal precision. On the other hand, existing mathematical models have not been specifically designed by taking into consideration the way people communicate spatial relations.; This thesis developed a formalism and tested its suitability for the way people communicate spatial relationships. This was done by performing a metric refinement of the 9-intersection topological model and resulted in a set of metric parameters for spatial relations. The significance of metric parameters for particular spatial terms was determined using data obtained from human-subject testing. These results were also validated independently for a set of spatial terms in English.; Using the metric parameters, 59 English language spatial terms were calibrated and a Metric Table of Spatial Terms was created. This table provides the necessary information to answer GIS queries or to elicit the best spatial relations scene descriptors. The use of these parameters gives this formalism the flexibility of being calibrated for different environments. This flexibility, in turn, allows the formalism to be the basis for representing and transferring the semantics of spatial relations within the Spatial Data Transfer Standards (SDTS). Results of this research will be useful to database designers, GIS standards committees, researchers of fundamental spatial theories, and computational linguists.