A Bayesian Approach to Uncertainty Modeling in OWL Ontology

Dealing with uncertainty is crucial in ontology engineering tasks such as domain modeling, ontology reasoning, and concept mapping between ontologies. This paper presents our on-going research on modeling uncertainty in ontologies based on Bayesian networks (BN). This includes 1) extending OWL to allow additional probabilistic markups for attaching probability information, 2) directly converting a probabilistically annotated OWL ontology into a BN structure by a set of structural translation rules, and 3) constructing the conditional probability tables (CPTs) of this BN using a new method based on iterative proportiobal fitting procedure (IPFP). The translated BN can support more accurate ontology reasoning under uncertainty as Bayesian inferences.