知识库的一致性问题和检查方法

Problems and Solving Methods on Knowledge Base Consistency

知识库中存在的不一致性是影响整个系统性能的重要因素之一。在系统建立初期,由于知识库比较小,内容也比较简单,只要设计人员或知识工程师对初始知识模型进行反复推敲和精心构造,这类问题还是比较容易防止的。但是随着时间的推移,人们对知识库在最初设计阶段的种种考虑逐渐淡漠,同时由于知识库中新的知识不断加入,知识库越来越大,内容也越来越丰富复杂,这时知识库中各知识单元之间的相互影响和相互联系就随之变得复杂,难以跟踪和捉摸了。在这种情况下,知识库中的不一致性检测、识别和纠正显得更为重要。1、知识库的一致性问题知识库系统的组织按其结构可以划分为事实性知识(领域知识)和控制性知识。

There is urgent need for sophisticated tools to help experts deal with diffi- cult and indispensable task of knowledge acquisition.The problem of detecting inconsistencies in built knowledge base is especially crucial.As the knowledgc base enlarges,this problem should commonly be solved by classical logic or ot- her techniques.This paper analyses the definition and the notion of inconsiste- ncy,also presents several solving methods for such issue in production system (including classical logic,Petri Net and Truth Maintenance techniques)and their individual characteristics.