Inconsistency-tolerant reasoning with OWL DL

Abstract

The Web Ontology Language (OWL) is a family of description logic based ontology languages for the Semantic Web and gives well defined meaning to web accessible information and services. The study of inconsistency-tolerant reasoning with description logic knowledge bases is especially important for the Semantic Web since knowledge is not always prefect within it. An important challenge is strengthening the inference power of inconsistency-tolerant reasoning because it is normally impossible for paraconsistent logics to obey all important properties of inference together. This paper presents a non-classical DL called quasi-classical description logic (QCDL) to tolerate inconsistency in OWL DL which is a most important sublanguage of OWL supporting those users who want the maximum expressiveness while retaining computational completeness (i.e., all conclusions are guaranteed to be computable) and decidability (i.e., all computations terminate in finite time). Instead of blocking those inference rules, we validate them conditionally and partially, under which more useful information can still be inferred when inconsistency occurs. This new non-classical DL possesses several important properties as well as its paraconsistency in DL, but it does not bring any extra complexity in worst case. Finally, a transformation-based algorithm is proposed to reduce reasoning problems in QCDL to those in DL so that existing OWL DL reasoners can be used to implement inconsistency-tolerant reasoning. Based on this algorithm, a prototype OWL DL paraconsistent reasoner called PROSE is implemented. Preliminary experiments shows that PROSE produces more intuitive results for inconsistent knowledge bases than other systems in general.