Semantic similarity applied to generalization of geospatial data

The paper presents an approach to verifying the consistency of generalized geospatial data at a conceptual level. The principal stages of the proposed methodology are Analysis, Synthesis, and Verification. Analysis is focused on extracting the peculiarities of spatial relations by means of quantitative measures. Synthesis is used to generate a conceptual representation (ontology) that explicitly and qualitatively represents the relations between geospatial objects, resulting in tuples called herein semantic descriptions. Verification consists of a comparison between two semantic descriptions (description of source and generalized data): we measure the semantic distance (confusion) between ontology local concepts, generating three global concepts Equal, Unequal, and Equivalent. They measure the (in) consistency of generalized data: Equal and Equivalent - their consistency, while Unequal - an inconsistency. The method does not depend on coordinates, scales, units of measure, cartographic projection, representation format, geometric primitives, and so on. The approach is applied and tested on the generalization of two topographic layers: rivers and elevation contour lines (case of study).