The Role of Crisp Elements in Fuzzy Ontologies: The Case of Fuzzy OWL 2 EL
Fernando Bobillo · IEEE Transactions on Fuzzy Systems · 2015
In the last years, fuzzy ontologies and fuzzy description logics (DLs) have been applied successfully to several real-world domains. Some recent results on fuzzy DLs show an increase in the theoretical complexity, even for not expressive languages. However, it still may be possible to obtain reasoning algorithms performing well in practice. This paper studies some cases where reasoning with a fuzzy ontology can be reduced to reasoning with a tractable classical language (an OWL 2 profile). Then, building on the assumption that not all concepts and roles in fuzzy ontologies are indeed fuzzy, we show that reasoning algorithms can be optimized by taking into account the existence of possible crisp concepts and roles. We also prove the usefulness on crisp representation reasoning algorithms over an empirical evaluation over fuzzy OWL 2 EL ontologies, and discuss some preliminary techniques to learn which elements should be considered to be crisp or noncrisp.