Handling uncertainty in ontology construction based on Bayesian approaches: A comparative study

Abstract

Ontology is widely used to represent knowledge in many software applications. By default, ontology languages such as OWL and RDF is built on discrete logic, so that it can not handle uncertain information about a domain. Various approaches have been made to represent uncertainty in ontology, one of which with a Bayesian approach. Currently, there are four published approaches: BayesOWL or OntoBayes, Multi-Entity Bayesian Networks (MEBN), Probabil istic OWL (PR-OWL), and Dempster-Shafer Theory. This paper provides a comparative study on those approaches based on complexity, accuracy, ease of implementation, reasoning, and tools support. The study concluded that Baye-sOWL is the most recommended approach to handle uncertainty in ontology construction among others.