Ontology Matching Using Multi-head Attention Graph Isomorphism Network

Ontology matching is a widely used solution to the semantic heterogeneity problem in data integration or sharing. It consists of establishing mappings between entities that belong to different ontologies. However, as the number of ontologies is increasing for a given domain and the overlap between ontologies grows proportionally, it becomes crucial to develop more reliable and accurate techniques for the automation of this task. While traditional ontology mapping approaches are based on string metrics and structure analysis, some recent methods are using deep neural networks. In this article, we propose a novel approach for ontology matching based on Graph Neural Networks (GNN) as graph representations are helpful for entity and graph comparisons. Our approach is more precisely based on Multi-Head Attention Graph Isomorphism Network (MHAGIN). The results of experiments demonstrate the effectiveness of our approach compared with existing methods.