A Comparative Study of Two Complex Ontologies in Air Traffic Management

Over the past 20 years, the ontology alignment community has developed many different matching algorithms for performing alignments. Each of these algorithms produces a proposed alignment between two input ontologies, which must be further scrutinized and validated prior to usage. In this paper, we examine an algorithm-independent approach intended to assist with the validation process. In particular, we have explored and tested a set of mismatch detection techniques that can identify unlikely matches within a generated alignment. We developed these methods based on our experience producing a reference alignment for two complex ontologies in the Air Traffic Management domain and subsequently observing poor performance when aligning these ontologies using state-of-the-art ontology matching systems. Our techniques are evaluated in a dataset which could serve as an interesting and challenging benchmark for the ontology matching community. Results from the evaluation show that ontology matching systems can benefit from such techniques as they contribute to a considerable increase in precision. Keywords: Ontolgies, Ontology Matching, SWIM