A Study on Local Search Meta-heuristics for Ontology Alignment

Abstract

Ontologies are one of the most suitable methodologies to provide formalized representations of the real-world data in several contexts such as the emerging paradigm of the Internet of Things. In spite of their key capability of providing an abstract representation of the information captured by different sources, the variety of ways that a domain can be conceptualized results in the development of heterogeneous ontologies with overlapping parts. In order to address this problem, a so-called ontology alignment process is required. This process allows generating a set of correspondences between semantically similar entities of two ontologies and, as a consequence, enabling system interoperability. Unfortunately, this process is a complex and time-consuming task. Therefore, recently, meta-heuristics are appearing as a suitable methodology to implement it. However, no meta-heuristics based on local search optimization have been applied until now. This paper bridges this gap by performing a comparison among some popular local search-based algorithms for generating an alignment between two ontologies. As shown by the results involving well-known benchmarks, Tabu search results to be the best performer in terms of precision, recall and F-measure.