Design And Implementation Of Mathematical Reasoning Engine Based On Graph Isomorphism

In recent years,artificial intelligence has gradually transitioned from traditional per-ceptual intelligence to cognitive intelligence,cognitive intelligence and automatic reason-ing have become the focus of research.how to apply deep learning to logical reasoning so that machines have the ability to think and reason will be a major breakthrough in artificial intelligence.The research content of this article is Design and construction of elementary mathematical reasoning engine based on graph isomorphism,The system design concept is based on a production system and involves two parts: knowledge representation and the construction of an instantiated rule base.The specific research content is as follows:(1)Knowledge representation of elementary mathematicsKnowledge representation is the first step for human-like solving system.Only when human semantics are converted into computer knowledge structures can subsequent rea-soning and Problem solving work.In this paper,the knowledge graph is finally selected to represent the conceptual entities in mathematics and the relationship between them.In the actual code,the class structure in Java is used to construct the entity classes and relation-ship classes of elementary mathematics.The inheritance idea of abstract classes is used to satisfy the knowledge graph.The principle of extensibility and opening and closing is that all entity classes inherit from abstract entity classes,and all relation classes inherit from abstract relation classes.(2)Build an instantiated theorem libraryIn the reasoning engine based on the production system,the rule base is an important external drive and the basis for generating knowledge.The rule base constructed in this paper is also expressed and stored in the form of knowledge graph.The construction of the instantiated theorem library is divided into three steps: theorem collection,theorem standardization,and theorem instantiation.The source of the theorem is the formula the-orem in the textbooks and teaching aids,the solution process of the standard answer,and the problem-solving skills.Theorem standardization is to convert the collected theorems into the standard format of the reasoning engine,which is one of the conditions for the simple and unified reasoning engine.Finally,the standardized theorem is generated into a knowledge graph to complete the instantiation.(3)Design and construction of reasoning engine of graph isomorphismIn the design of the reasoning engine in this paper,the forward inference method and the backward inference method in the conventional problem-solving thinking are in-tegrated to form a logical reasoning method of ”reverse first,then forward”,and it is combined with the computational reasoning method.In terms of architecture,a hierarchi-cal structure is adopted,and each logical layer maintains relative independence.The core algorithm is a graph matching method,which is a hybrid matching algorithm.The map-ping of entities and relationships is constructed through type matching on the knowledge graph,and then the string is pattern-matched to generate a symbol rotation set.This paper has completed the entire algorithm design of the mathematical reasoning engine based on graph isomorphism and the construction of all modules,and conducted random batch tests of non-application problems.The comprehensive problem solving rate is 71.2%,and the average solving time is no more than 5 minutes.