Using ontology evidences to extend belief network IR model

The retrieval performance of an information system usually increases when it uses the relationships among the terms. Belief network IR model provides a framework to combine deferent types of query evidences. In this paper we present an extended belief network model that combines user query terms and its ontology related terms. Then we evaluate the extended model using a reference collection of 1359 documents. The results show that the new model has improved retrieval performance.