To address semantic heterogeneity, ontology matching finds correspondences between entities from different representations of knowledge from the same domain. Most existing ontology matching systems lack the ability to provide explanations for their results. In this paper, we propose an abduction-based approach to explain generated correspondences by existing ontology matching systems. The method takes as input two ontologies and a set of aligned entities and generates explanations for the established relations via abduction by utilizing isomorphisms between graphical representations of entity definitions. We enhance existing ontology matching systems by providing an out-of-the-box component for generating explanations of ontology matching results. We present a formal framework for abduction-based explanations of ontology matches and illustrate our approach using an example from the alignment of two medical ontologies from the Bio-ML track of the Ontology Alignment Evaluation Initiative (OAEI). Further, we conduct experiments on generating explanations for aligned ontologies from the Bio-ML track and report on the results.

错误:搜索内容不能为空,请输入英文关键词
错误:关键词超出字数限制,请精简
高级检索

Explaining Ontology Alignments via Abductive Graph Isomorphisms

  • Ivan Gocev,
  • Georgios Meditskos,
  • Nick Bassiliades

摘要

To address semantic heterogeneity, ontology matching finds correspondences between entities from different representations of knowledge from the same domain. Most existing ontology matching systems lack the ability to provide explanations for their results. In this paper, we propose an abduction-based approach to explain generated correspondences by existing ontology matching systems. The method takes as input two ontologies and a set of aligned entities and generates explanations for the established relations via abduction by utilizing isomorphisms between graphical representations of entity definitions. We enhance existing ontology matching systems by providing an out-of-the-box component for generating explanations of ontology matching results. We present a formal framework for abduction-based explanations of ontology matches and illustrate our approach using an example from the alignment of two medical ontologies from the Bio-ML track of the Ontology Alignment Evaluation Initiative (OAEI). Further, we conduct experiments on generating explanations for aligned ontologies from the Bio-ML track and report on the results.