Towards Efficient Axiom Pinpointing of EL+ Ontologies

Abstract: The EL family of Description Logics (DLs) has been the subject of interest in recent years. On the one hand, these DLs are tractable, but fairly inexpressive. On the other hand, these DLs can be used for designing different classes of ontologies, most notably ontologies from the medical domain. Unfortunately, building ontologies is error-prone. As a result, inferable subsumption relations among concepts may be unintended. In recent years, the problem of axiom pinpointing has been studied with the purpose of providing minimal sets of axioms that explain unintended subsumption relations. For the concrete case of EL and EL+, the most efficient approaches consist of encoding the problem into propositional logic, specifically as a Horn formula, which is then analyzed with a dedicated algorithm. This paper builds on this earlier work, but exploits the important relationship between minimal axioms sets and minimal unsatisfiable subformulas in the propositional domain. In turn, this relationship allows applying a vast body of recent work in the propositional domain to the concrete case of axiom pinpointing for EL and its variants. From a practical perspective, the algorithms described in this paper are often several orders of magnitude more efficient that the current state of the art in axiom pinpointing for the EL family of DLs.