Finding Similar Deductive Consequences — A New Search-Based Framework for Unified Reasoning from Cases and General Knowledge

摘要

While reasoning with cases is usually done in a similarity-based manner, additional general knowledge is often represented in rules, constraints, or ontology definitions and is applied in a deductive reasoning process. This paper presents a new view on the combination of deductive and similarity-based reasoning, which is embedded in the CBR context. The basic idea is to view general knowledge and cases as a logical theory of a domain. Similarity-based reasoning is introduced as search for the most similar element in the deductive closure of the domain theory. We elaborate this approach and introduce several related search algorithms, which are analyzed in an experimental study. Further, we show how several previous approaches for using general knowledge in CBR can be mapped to our new view.