Research On The Construction Of Knowledge Graph And Classification Algorithm Of Math Word Problem

The automatic solution of mathematical word problems(MWP)has always been the difficulty and focus in the field of machine intelligence research.As early as 1960s,some scholars have devoted themselves to the research in this field.In recent years,with the rapid development of machine learning,a large number of researchers had solved mathe-matical word problems through new technologies.Solving this problem requires a variety of techniques,including converting human language descriptions into machine-readable sentences that machines can use to compute the correct answer.It is a comprehensive problem combining natural language understanding and automatic reasoning.In this pa-per,a method of classifying and solving problems is chosen to solve math word problems automatically.The main research contents of this paper are as follows:(1)In this article we builds a knowledge map of math word problem,this example because the knowledge map of research is more and more attention in recent years,using the relevant technology of the knowledge map construction,used to solve the relations among entities which need to extract the problem solving,build the map data mainly from corpus,this example entity extraction through the corpus,the nature of the same entity into one entity type,and then define the existence of the relationship between the entity type.The entity classes and relation classes are constructed by Java language.With Neo4j graph database is connected,and these entity classes and relation classes are saved as entity nodes and relation edges in the graph database.The form of nodes and edges of the Neo4j graph database also completes data visualization.We have now completed application of types of knowledge map with 103 nodes and 134 relationship,other types of knowledge map in the building,entities extracted from text are extracted through the knowledge graph for entity type extraction,and then whether there is a relationship between the entity types in the knowledge graph is queried,and relationship labeling is carried out to obtain the information needed for problem solving.(2)The construction of word problem classification system and the research of auto-matic classification algorithm.By analyzing the data of word problems and centering on the mathematics teaching system of middle schools in China,the first-level classification standard of word problems is constructed based on the knowledge points and the problem solving model through the solution process and solving ideas,so as to avoid the inaccuracy of the classification caused by subjective deviation.Each category is classified into two-level,and the detailed classification system of word problems is completed.Then,the text classification technology is applied to the automatic classification of word problems.The research mainly includes the extraction and representation of features,and the training of classification model.The classification effect of traditional machine learning classifica-tion and deep learning classification model is compared,and the final classification effect of word problems is 84.2%.