Similarity from conceptual relations

Abstract

The main focus of this paper is how to measure similarity in a content-based information retrieval environment. In the first part we define the information base, which is a generative framework where an ontology in combination with a concept language defines a set of well-formed concepts. Well-formed concepts is assumed to be the basis for an indexing of the information base in the sense that these concepts appear in descriptions attached to objects in the base. Subsequent and last we introduce an approach for measuring similarity in this framework. The measuring problem is divided into to continuous parts where we first narrow what concepts have in common, and secondly use this fragment, a similarity graph, for calculating the similarity between concepts. The purpose of narrowing or restricting what concepts have in common is to manage the generative aspect of the ontology, and to retain the greatest possible number of shared attributes and characteristics of the concepts being compared. Taking the similarity graphs as input we discuss what properties a similarity function need to satisfy to measure the degree of similarity proportional to how close the concepts are or how much they share.