Presenting a Method for Aligning Ontologies Based on Structural Graphs

Abstract

The need to give meaning to web content in a way that is understandable to different machines is the main idea behind the emergence of the Semantic Web, which is based on ontologies, as structures that model data in the form of words and concepts. In this paper, we present a method for aligning existing ontologies based on recognizing lexical, semantic, and structural similarities between entities in these ontologies. To identify mappings related to two ontologies automatically, the first and most important step is to identify anchors. All alignment calculations are performed based on the initial anchors. In the proposed method, anchors are identified by lexical and semantic analysis of entities. In the lexical matching step, entity labels are compared based on five criteria. To integrate the effect of the lexical similarity criteria set, averaging of normal values ​​is used. For semantic matching of two ontologies, the Verdant tool is also used. In the structural matching section of ontologies, the similarity between entities is recalculated based on the ontology graph and the similarity of the neighborhoods of the entities. In addition, the feature matching operation is also considered. In the last step, the resulting mapping set is filtered and a 1-1 mapping is presented as the final alignment set. To compare and evaluate the proposed method, the OAEI dataset is used and the results of this method are compared with other proposed systems using the evaluation criteria, precision, recall, and F-criterion. The evaluations performed show the efficiency of the proposed system compared to other systems under consideration.