Representation of Concept Specialization Distance through Resemblance Relations (2003) [5 citations — 4 self]
Abstract:
Summary. Generalization-specialization (gen-spec) relationships between pairs of classifiers can be assigned a grade of strength or relative distance, somewhat representing the level of similarity between the class and its subclass. In many cases, these distances can not be adequately computed from the structural features or properties of the classifiers, since class-subclass discrimination semantics are often not represented explicitly, and distances can only be properly assessed subjectively by humans. In this work, we describe a novel approach that uses resemblance relations to model graded specializations, both from a specific classifier (locally) and also along a subset of the generalization hierarchy (globally). We also show how that approach can be combined with current Web-enabled ontology description languages to carry out adaptive behaviors that involve crawling the gen-spec hierarchy. Key words: Generalization/specialization relationship, subtyping, ontology, resemblance relations, RDF
