Reverse Ontology Matching Jorge Martinez-Gil University of Malaga Dept. of Computing Sciences Boulevard Louis Pasteur 35, 29071 Malaga jorgemar@lcc.uma.es Jose F. Aldana-Montes University of Malaga Dept. of Computing Sciences Boulevard Louis Pasteur 35, 29071 Malaga jfam@lcc.uma.es ABSTRACT Ontology Matching aims to find the semantic correspon- dences between ontologies that belong to a single do- main but that have been developed separately. How- ever, there are still some problem areas to be solved, be- cause experts are still needed to supervise the matching processes and an efficient way to reuse the alignments has not yet been found. We propose a novel technique named Reverse Ontology Matching, which aims to find the matching functions that were used in the original process. The use of these functions is very useful for aspects such as modeling behavior from experts, per- forming matching-by-example, reverse engineering ex- isting ontology matching tools or compressing ontology alignment repositories. Moreover, the results obtained from a widely used benchmark dataset provide evidence of the effectiveness of this approach. 1. INTRODUCTION In the new approaches to develop information systems, the use of a type of formal schema called ontology is usual. Ontologies are considered to be semantically richer than schemas in general, and therefore, techniques for schema matching can be easily adapted to ontologies but not vice versa [12]. There are many ontologies available on the web currently. These ontologies are usually developed for different collections of information, and differ- ent kinds of applications. Nowadays, the Swoogle search engine 1 has indexed thousands of ontologies. There are several reasons for the quick proliferation of ontologies, but we consider mainly two: • It is often easier to construct a new ontology, than find an existing one which is appropriate for a given task. • There is often a desire for direct control over the ontology for a particular domain, rather than having the structure dictated by external forces. 1 http://swoogle.umbc.edu A direct consequence of having large numbers of ontologies available is that it is necessary to inte- grate knowledge which is represented in different ways. Ontology matching aims to produce align- ments, that is, sets of semantic correspondences be- tween elements from different ontologies. This task is very expensive in terms of time and resource con- sumption. The reason is that it is necessary a lot of work from domain experts to match ontologies or to supervise results from existing semiautomatic tools. Our approach is based on the extraction of the ontology matching functions used by the agents, experts or tools when matching ontologies, so it is a powerful way to reuse, store and understand their knowledge. Moreover, there are other collat- eral benefits as the ability to implement strategies for ontology matching by example, reverse engineer- ing existing ontology matching tools or compress large ontology alignment repositories. In this way, we think that the main contributions of our work can be summarized as follow: • We propose, for the first time to the best of our knowledge, a methodology for reverse en- gineering an ontology alignment which tries to find the matching function that have been used to generate an ontology alignment. • We perform an empirical evaluation of our ap- proach in order to show its practical viability in the real world. The rest of this work is structured in the follow- ing way: Section 2 describes the problem statement related to Reverse Ontology Matching. Section 3 presents the related works regarding other reverse engineering proposals. Section 4 presents the core of our approach, a methodology for reverse engi- neer an ontology alignment, and some real exam- ples. Section 5 contains an evaluation that shows the applicability of Reverse Ontology Matching in the practice. In Section 6, we describe the conclu- sions extracted from this work. SIGMOD Record, December 2010 (Vol. 39, No. 4) 5