Coincidence-Based Scoring of Mappings in Ontology Alignment

Abstract

Ontology Matching (OM) which targets finding a set of alignments across two ontologies, is a key enabler for the success of Semantic Web. In this paper, we introduce a new perspective on this problem. By interpreting ontologies as Typed Graphs embedded in a Metric Space, coincidence of the structures of the two ontologies is formulated. Having such a formulation, we define a mechanism to score mappings. This scoring can then be used to extract a good alignment among a number of candidates. To do this, this paper introduces three approaches: The first one, straightforward and capable of finding the optimum alignment, investigates all possible alignments, but its runtime complexity limits its use to small ontologies only. To overcome this shortcoming, we introduce a second solution as well which employs a Genetic Algorithm (GA) and shows a good effectiveness for some certain test collections. Based on approximative approaches, a third solution is also provided which, for the same purpose, measures random walks in each ontology versus the other.