AN ANALYSIS OF MULTIPLE SIMILARITY MEASURES FOR ONTOLOGY MAPPING PROBLEM

Abstract

This paper presents an analysis of similarity measures for the ontology mapping problem. To that end, 48 similarity measures such as string matching and knowledge based similarities that have been widely used in ontology mapping systems are defined. The similarity measures are investigated by discriminant analysis with a real-world data set. As a result, it was possible to identify 22 effective similarity measures for the ontology mapping problem out of 48 possible similarity measures. The identified measures have a wide variety in the type of similarity. To test whether the identified similarity measures are effective for the problem, experiments were conducted with all 48 similarity measures and the 22 identified similarity measures by using two major machine learning methods, decision tree and support vector machine. The experimental results show that the performance of the 48 cases and the 22 cases is almost the same regardless of the machine learning method. This implies that effective features for the ontology mapping problem were successfully identified.