| Abstract: | Ontology occupies an important position in artificial intelligence, computer linguistics and knowledge management. However, when different ontologies are constructed to represent the same information in a domain, the so-called heterogeneity problem arises. In order to address this problem, a key task is to discover the semantic relationship of entities between given two ontologies, called ontology alignment. Recently, the meta-heuristic algorithms have already been regarded as an effective approach for solving ontology alignment problem. However, firstly, as the ontologies become increasingly large, meta-heuristic algorithms may be easier to find local optimal alignment in large search spaces. Secondly, many existing approaches exploit the population-based meta-heuristic algorithms so that the massive calculation is required. In this paper, an improved compact particle swarm algorithm by using a local search strategy is proposed, called LSCPSOA, to improve the performance of finding more correct correspondences. In LSCPSOA, two update strategies with local search capability are employed to avoid falling into a local optimal alignment. The proposed algorithm has been evaluated on several large ontology data sets and compared with existing ontology alignment methods. The experimental results show that the proposed algorithm can find more correct correspondences and improves the time performance compared with other meta-heuristic algorithms. |
