Research On Entity Relation Extraction Method For Machine Understanding Of Elementary Mathematics Problems

With the continuous development of artificial intelligence,the application of intelligent question answering,machine proving and automatic solving based on artificial intelligence technology has become a research hotspot,and has made new breakthroughs.Understanding the meaning of mathematical problems is undoubtedly one of the most attractive research topics in the field of artificial intelligence.The goal of topic understanding is to realize the formal representation of mathematical problems,which needs to transform the unstructured mathematical language text into structured data.Entity relation extraction,as an important part and core task in the fields of information extraction,automatic question answering and text summarization,is also the core part in the process of understanding the meaning of elementary mathematical problems.The semantics of elementary mathematics problems are complex and the context is changeable,which undoubtedly adds difficulty to the task of entity relation extraction.In addition,there are some difficulties in elementary mathematics,such as overlapping entity relationship recognition and multi entity relationship extraction across sentences.In view of the above problems,this study systematically analyzes the characteristics of elementary mathematics problem language,and on this basis,takes the classical probability word problem in elementary mathematics as the research object to extract entity relationship.The specific research contents include the following four aspects.(1)Based on the difficulty analysis of text features and entity relation extraction of elementary mathematics problems,this paper proposes a annotation method for entity relation extraction of elementary mathematics problems,hoping that the method can effectively identify the relationship between adjacent entities,and lay the foundation for the subsequent task of entity relation extraction of the whole problem.(2)Based on the proposed annotation method,this paper takes the classical probability word problem in elementary mathematics as an example.In this study,the bilstm-crf model is used to test 939 classical probability word problem data sets,and the optimal F1 value of the model is 89.75%.At the same time,the results are compared with those of CRF model.The advantages and disadvantages of the two models are analyzed,and the effectiveness of the annotation method and the model is verified.(3)Based on the entity relationship triples of machine recognition,this study constructs the relationship reasoning rules,transforms the existing entity relationship into the relationship matrix,and uses the depth first search algorithm to mine the hidden relationship,so as to complete the entity relationship extraction of the whole question.(4)Based on the proposed entity relation extraction method for elementary mathematics problems and the established entity relation reasoning rules,this paper designs and develops an entity relation extraction system for classical probability word problem based on the existing data sets.The system can realize text input,preprocessing,entity recognition,entity relation extraction,entity relation network generation,entity relation recognition,entity relation recognition,entity relation recognition,entity relation recognition and entity relation network generation The purpose of this paper is to revise the entity relation and save the subject information and entity relation,so as to help the subsequent research on the understanding of the subject meaning of elementary mathematics problems.