I. Gocev, G. Meditskos, N. Bassiliades, “Towards Abduction-based Ontology Matching”, accepted, 16th IEEE International Conference on Knowledge Graphs (ICKG 2025), November 13-14, 2025, Limassol, Cyprus.
The purpose of this paper is to present ongoing research on ontology matching, with a specific focus on exploring the use of abductive reasoning as a core mechanism. In particular, we aim to investigate how abduction can be leveraged to address common challenges in ontology matching, such as (automatic) discovering and explaining of simple and complex correspondences, as well as address scalability issues in ontology matching. Moreover, explainability in ontology matching is much needed and there do not exist many formal frameworks for this. We found that depicting entity relations as observations offers the possibility to discover missing correspondences by using abduction, which can be used to perform and explain simple and complex ontology alignments. In addition, we identified certain limitations that come as a result of using abduction, such as scalability, which is a common challenge in ontology matching. To perform ontology matching using abduction, we opt to search for isomorphisms between graphical representations of ontologies and concept definitions within them. Considering that structural preserving maps characterize entity relations, this would provide the opportunity to identify existing and find missing relations between entities in different ontologies and align them. In
addition, a heuristic-based ranking of abductive solutions is introduced, to address the scalability of our approach. To this end, we present a formal definition for abduction-based ontology matching and illustrate how our approach can be used to achieve and explain ontology alignments. This provides transparency to the matching process and helps users better understand ontology alignments. We introduce a novel heuristic-based ranking of semantic matches, i.e. abductive solutions, and, consequently, of ontology alignments. Furthermore, we conduct experiments on real-world ontologies used in the Ontology Alignment Evaluation Initiative (OAEI) and report on the results.
