OWL_strict: A Constrained OWL Fragment to Avoid Ambiguities for Knowledge Graph Practitioners

Abstract

While RDF is a standard exchange format for directed labelled graphs, the Web Ontology Language (OWL) is a likewise standardized, powerful and flexible language to express terminological axioms for describing the schema of RDF graphs. As such, RDF and OWL are natural choices for interchanging and describing the schema of Knowledge Graphs (KGs) in practice. However, the syntactic “freedom” of RDF can lead to unintuitive OWL ontologies that are challenging for implementers of KGs to use and which can cause ambiguous interpretations when used in ontology editors, or – even worse – hint at modeling errors. In this paper, starting from the notion of “standard-use” of RDF, we aim at providing a methodology and tool support for the challenge of solving respective ambiguity problems in the context of ontology editors. We first define a syntactically constrained fragment of OWL, OWLstrict, which aims to avoid ambiguities based on requirements identified in ontology editors in practice.To strengthen the argument for avoiding these ambiguities, we compare how different ontology editors handle them. Second, we discuss how far this fragment can be syntactically described and checked using SHACL.Third, to provide tool support and building upon our prior work on SHACL repairs, we discuss how non-OWLstrict-compliant ontologies can be repaired, i.e., modified to be made compliant with OWLstrict. We perform an evaluation with a particular ontology editor and six ontologies we use in real-world projects and show that the repaired versions satisfy OWLstrict conformance requirements and are therefore compatible with the editor.