Extending ontology queries with Bayesian network reasoning

Abstract

The paper presents an approach for adding probabilistic reasoning to ontologies. We show how an extended form of Bayesian network can be extracted from an existing ontology. The structure of the Bayesian Network is obtained by means of an analysis of the classes and their relationships, and the initial distributions are obtained by considering the instances. The resulting Bayesian network offers reasoning capabilities that can satisfy an interesting set of probabilistic queries.