Two-sided hypotheses generation for abductive analogical reasoning

Abstract

In general, if a knowledge base lacks the necessary knowledge, abductive reasoning cannot explain an observation. Therefore, it is necessary to generate missing hypotheses. CMS can generate missing hypotheses, but it can only generate short-cut hypotheses or hypotheses that will not be placed on real leaves. That is, the inference path is incomplete (truncated), so that abduction is not complete. The inference proposed tries to generate missing hypotheses that are placed on the middle of the inference path by both abductive inference and deductive inference using analogical mapping. As a result, the inference can generate missing hypotheses even on the middle of the inference path.