一种基于未命名概念的类比推理方法

An Analogical Reasoning Method Based on the Unnamed Concept

类比推理作为人类认知发展的中心能力之一,是人工智能领域的一个研究重点.传统的类比推理方法只以人类所熟知的命名概念为基础进行类比推理,忽略了大规模数据中往往存在许多尚未认知的"准概念"——未命名概念,限制了类比推理的完整性.针对这一问题,结合概念格在存储数据信息方面的优越性和知识表达的完备性,提出一种基于未命名概念的类比推理方法.该方法在类比推理的过程中以概念格模型为核心,首先建立源领域和靶向领域的概念格模型,通过概念格结构将两个领域的数据组织起来,其次利用同构算法得到源和靶域之间的节点映射,然后通过本体和概念的属性匹配得到未命名概念,最后利用未命名概念将源领域中的映射关系类比至靶领域,扩大类比结果.实例分析结果表明,文章所提出的基于未命名概念的类比推理方法能够提高类比推理的完整性,得到更多的类比结果.

An analogical reasoning is one of important abilities about human cognitive development, and it has become a popular issue in artificial intelligence. Traditional analogical reasoning methods usually utilize named concept known by common people to reasoning, but these methods ignore that large scale data often existed many ＂ quasi concept＂ which is unnamed concept, thus limiting the integrity of analogical reasoning. To focus this problem, combining the advantages of concept lattice in storing data information and knowledge representation, this paper proposed an analogical reasoning method by unnamed concept on the basis of concept lattice as core data structure. Firstly, in the process of analogy reasoning, we use the concept lattice model to establish the concept relations of the source domain and the target domain, and organize the data of two domains by the concept lattice structure, Secondly, we acquire the nodes mapping relations between the source and the target by exploiting the isomorphic algorithm. Then, we obtain the unnamed concepts by attributing matching of the concept of ontology and concept nodes. Finally, we utilize those unnamed concepts to make analogical reasoning from source domain to target domain by structure mapping relation, and the analogy results are expanded. The experimental results show that the method of analogical reasoning based on unnamed concept proposed by this paper can improve the integrity of analogical reasoning, and get more analogy results.