A Method for Judging Illogical Discourse Based on Concept Association and Common-Sense Judgment

摘要

In recent years, research into natural language processing for implementing natural discourse has attracted considerable attention. However, up to now, research into natural language processing has focused mainly on discourse processing for certain limited objectives and under specific conditions, with emphasis on the superficial formation of sentences. The understanding of the user-generated words is constrained by the size of the constructed knowledge base and the quality of actual data collected. Overcoming this limitation requires the capability to autonomously generate discourse sentences using a computer. However, there is a risk that the generation of response sentences by a computer results in illogical word combinations caused by the mechanical combination of words. To avoid this risk, the capability of detecting illogical word combinations is needed. This paper proposes a technique for judging illogical word combinations by creating a knowledge model of human discourse and words.