Research On Ontology Uncertainty Reasoning Based On Bayesian Network

Ontology is considered as one of the pillars of the Semantic Web. It is a modeling tool for knowledge presentation both in semantic and knowledge hierarchy. It can provide conceptual specific description and build the foundation for knowledge sharing. However, Ontology can not express the overlap or the degree of intersection, and can not support uncertainty reasoning. Bayesian decision theory provided the theory foundation of uncertainty event or reason. In uncertainty knowledge presentation and reasoning, Bayesian Network proved to be one of most effective methods in obtaining belief of uncertainty knowledge. This paper integrated Bayesian Network in uncertainty reasoning, extended OWL with probability description, so as to let it support uncertainty knowledge and incomplete or imprecise information, thus we can preserve the advantages of both, for ontology has power of knowledge presenting and Bayesian network provides ability of reasoning. In this way, we extend the ability of uncertainty presentation and reasoning in Ontology, and can make Ontology widely used.Firstly,we introduced the foundation of Ontology briefly. Ontology is an explicit, formal representation of the entities and relationships that can exist in a domain of application. It can provides the common understanding of a domain, and provides the explicit relationship between the terms in different hierarchy. We extend web ontology language (OWL) with probability information, make it support uncertainty presentation, and we build a domain ontology which was extended by probability information in Protégé3.3.Secondly, we also make research on the Bayesian Network. Bayesian Network is an integration of Graph Theory and Probability Theory. It indicates the Conditional Independence and the Probability Distribution among Variables and is wildly used in probability reasoning, for there are uncertainties in our real world and Bayesian Network can present the uncertainty in a modular way. We also introduced several other uncertainty reasoning methods and their features.Finally,we developed a framework named OntoBN to translate the probability ontology into a Bayesian Network. It can extract the probability information from ontology and translate the probability concepts into nodes in Bayesian Network, and can reason by the Bayesian Network. The experiment result shows that we can gain more information from it, and can provide useful information for our decision.