Explaining Answers from Agent Communication of Semantic Web Information

Abstract

In semantic Web, an application can not only provide explicit information but also infer implicit knowledge. Besides, an answer can be also derived from multiple sources and by multiple agents. In order to trust and use such an answer, a user needs a proof and/or explanations of that answer, which contain information related to knowledge sources, systems, reasoning methods, relationships of data items, etc. In this paper, we extend our existing logical agent framework so that agents can not just reason with multiple distributed ontologies but also record proofs and provide explanations of their answers. A proof contains essential information related to a reasoning process including agents, sources of ontologies, rules, facts, etc. Then various explanations can be then generated from a proof for users, depending on their purpose and requirement. Agents are capable of creating, exchanging, and accumulating (sub-) answers along with their proofs and explanations among each other when they cooperate in solving common problems.