| In recent years,the combination of traditional education and cutting-edge artificial intelligence technology has received more and more attention.As one of the important basic subjects in the field of education,elementary mathematics,combined with natural language processing and other technologies,to realize the automatic solution of elementary mathematics problems is a national priority The research directions supported by national key projects are of great significance in the field of education and teaching.Automatically solving elementary mathematics problems is a challenging topic,which mainly includes two parts: natural language understanding of elementary mathematics and automatic reasoning.As the upstream task of automatic problem-solving reasoning,natural language understanding of mathematics needs to ensure that elementary mathematics texts are converted into structured form that computer can process them.It lays the foundation for subsequent reasoning and problem-solving tasks.Therefore,how to accurately and completely extract valuable information such as entities and relations in elementary mathematics texts is a major research problem.The main research content of the thesis is as follows:1)Based on related technologies such as knowledge graph and natural language processing,combined with the characteristics of elementary mathematics language and knowledge,the thesis designs an elementary mathematics knowledge graph.2)Further,conduct in-depth research on the implicit entities and implicit relationships in elementary mathematics texts,and propose an automatic implicit entity extraction algorithm based on the combination of phrase structure tree analysis and dependency syntax analysis.The concept knowledge graph of elementary mathematics is used as benchmark for automatic extraction of implicit entities and relationships to standardize entities and relationships.3)On this basis,the thesis constructs an automatic implicit entities and relationships extraction module,and experiments are carried out on the dataset of elementary mathematics topics,and the accuracy rate of implicit triples extraction can reach 91.38%.Experiments have proved that this method can more accurately discover the implicit entities and relationships in the elementary mathematics text.It can ensure the integrity of the transformation of elementary mathematics natural language into structured information,and better provide intelligent support for automatic problem solving. |
