The aim of this study is to develop a commonsense judgment mechanism for intelligent information processing such as a human being does. The principal element of the commonsense judgment mechanism is a concept-base defining concepts by attributes and their weights. It is expected that the concept-base is automatically constructed. The description of concepts in the concept-base constructed automatically, however, is inferior to that well prepared by humans in quality. To cope with this situation, this paper proposes a mechanical refinement method of the concept-base. The method uses a degree of association among concepts, the thesaurus and so forth. It is shown that the proposed method refines the concept-base and yields a degree of association closer to that decided by human judgment.
展开▼