Personalized Reasoner Based on Belief Strengths of Information Sources

Abstract

Traditionally, only “individual-independent ” facts are inputted to a reasoner. The reasoner will produce the same answer to the same question, regardless of who the questioner is. However, since ontology reasoners have their facts distributed in the open internet environment, reasoners may contain “individual-dependent ” facts. Personalized reasoner is becoming possible. Furthermore, gathering “individual-dependent ” facts from different individuals often contains inconsistent information or contradictions (Staab 2004). Manually solving the contradictions as usual is not a practical way out. Capability for personalized reasoner to reason with inconsistency should be developed. A personalized reasoner based on belief strengths of information sources is proposed in the paper. Key steps toward such a personalized reasoner are outlined below. 1. Identify information providers and questioners, who are persons (or organizations) in most cases. Ontologies about person (or organization) has been constructed and widely used. It’s easy to adopt some of these ontologies to describe providers and questioners. 2. Calculate questioners ’ belief strengths of information sources based on identification information of questioners and providers. Semantic distance function can be designed to calculate the distance between provider and questioner to a value in (0..1], which is then mapped to di in BL stated below. 3. Reason with facts with belief strengths. This is a novel extension to the current DLs and some details are given below.