Inference in a Topically Organized Semantic Net

A semantic net system in which knowled; ge is topically organized around concepts has been under development at the University of Alberta for some time. The system is capable of automatic topical classification of modal logic input sentences, concept and topic oriented retrieval, and property inheritance of a general sort. This paper presents an inference method which efficiently determines yes or no answers to relatively simple questions about knowledge in the net. It is a deductive, resolution based method, enhanced by a set of special inference methods, and relies on the classification and retrieval mechanisms of the net to maintain its effectiveness, unencumbered by the volume or diversity of knowledge in the net.