基于ALCQ(D)的CBR事例修正算法研究

Research on CBR Case Adaptation Based on ALCQ(D)

CBR(基于事例推理)是人工智能领域的一个分支,它克服了知识获取的瓶颈问题,事例修正是CBR的关键步骤。以ALC为代表的描述逻辑已被充分应用到CBR中,但目前在基于描述逻辑的CBR中还没有比较有效的算法来判断检索到的相似事例是否需要修正和如何进行修正。ALCQ(D)是在ALC的基础上引入定性数量约束Q和有型域D得到的。提出的算法用ALCQ(D)概念来描述CBR源事例和目标事例,先假定检索到的相似事例能够解决目标问题,即假定目标事例和相似事例同时满足知识库,但这样可能会与知识库产生冲突;接着使用冲突检测机制来查找相似事例概念描述中导致冲突的概念;最后使用概念替换规则在TBox本体库中检索该概念的最相似概念去替换它自己。研究表明,该算法具有界限性、可靠性和完备性。通过一个实例对其进行检验,结果表明,该算法可以准确修正检索到的相似事例,解决目标问题。

CBR（Case-based Reasoning）is a branch of artificial intelligence,which overcomes the bottleneck of knowledge acquisition.Case adaptation is a key step in CBR.Description logic ALC has been fully applied to the CBR,but now no alogirhtm is more effective to determine whether a retrieved similar case needs to be modified and how to fix based on description logic.ALC becomes ALCQ（D）as it introduces a qualitative quantity Q and a type constraint domain D.The algorithm in this paper uses ALCQ（D）concept to represent the source case and target case.Firstly,it presupposes that the retrieved source example can solve the problem of target,which means the target and source case examples both satisfy KB（knowledge base）,but this may lead to inconsistent with KB.Then according to the conflict detection in the algorithm,it finds the concepts which lead to inconsistent in source concept instances and finally uses concept of replacement rules defined in this article to retrieve the most similar concepts to the inconsistent concept in ontology repository for replacing itself.Studies show that this algorithm has boundaries,reliability and completeness.This paper used an example to illustrate this algorithm.The results show that it can revise the similar case to solve the target problem.