Research On The Construction And Representation Of Elementary Mathematics Knowledge Graph

With the rapid development of the Internet and big data,massive amounts of structured or unstructured data are being generated around the world at all times.These data formats are diversified,including pictures,text,audio,video,and so on.Due to the diversity of data and data representation,big data will not be effectively applied in many cases without structured processing.This article focuses on how to represent multi-source heterogeneous unstructured and semi-structured elementary mathematics data as structured data,and construct a knowledge graph of elementary mathematics concepts based on the method of constructing a knowledge graph of mathematics.Then apply the conceptual knowledge graph to the understanding of pure mathematics text,construct the elementary mathematics knowledge graph for reasoning system,and solve the basic problems of elementary mathematics.The storage form of the knowledge graph is stored in the form of triples.The reasoning algorithm applied to the graph neural network and the graph is not very suitable.The TransE algorithm is used to train the triples in the knowledge graph into word vectors,and the word vectors can be used in terms of predicting tail entities and graph reasoning based on head entities and relationships.In summary,this article mainly completed the following tasks:(1)Formulate the representation of elementary mathematics knowledge.Complete entity structure representation,relation structure representation,and knowledge characteristic representation in the field of mathematics,such as the representation of propositions,the mutual conversion of binary relations and multiple relations,etc.(2)Construct a knowledge graph of elementary mathematics concepts.The completed conceptual knowledge graph contains 551 entities,561 relations,and 204763 triples.(3)The triples in the conceptual knowledge graph is trained into word vectors through the graph embedding model.The word vector model can be used to predict the tail entities based on the head entity and the relationship.(4)Based on the knowledge graph of elementary mathematics concepts,combined with natural language understanding to generate a question knowledge graph,the question knowledge graph can be used in a reasoning system.The question knowledge graph generation system can be used for the conversion of any mathematic question text to knowledge graph.After testing 100 questions in function,sequence,geometry and other modules,the entity correct rate is above 90%,and the relationship correct rate is above 90%.