A neural model for ontology matching

Ontology matching is a key issue in the Semantic Web. The paper describes an unsupervised neural model for matching pairs of ontologies. The result of matching two ontologies is a class alignment, where each concept in one ontology is put into correspondence with a semantically related concept in the other one. The framework is based on a model of hierarchical self-organizing maps. Every concept of the two ontologies that are matched is encoded in a bag-of-words style, by counting the words that occur in their OWL concept definition. We evaluated this ontology matching model with the OAEI benchmark data set for the bibliography domain. For our experiments we chose pairs of ontologies from the dataset as candidates for matching.