Research And Application Of Entity Relation Automatic Extraction Algorithm In Elementary Mathematical Problems

With the rapid development of computer computing power and Internet,artificial intelligence ushered in the third wave of development.Under the active encouragement of the policy of strengthening the country with artificial intelligence,the artificial intelligence internet education has become the next outlet,which has given birth to many internet education enterprises.However,the Internet education has moved the offline classroom to the online classroom,which can not fundamentally meet the educational demands of the individualized education of the educated.Therefore,researchers began to explore whether the use of artificial intelligence technology to build a full-automatic teaching system can provide each educated with customized educational content.With its strict natural language style,mathematics has become the first pilot field in the development of intelligent education.Driven by this demand,it has become a hot spot for natural language researchers to explore whether computers have the ability to under-stand and deal with mathematical problems described by natural language.Using natural language processing technology to make computer understand mathematical natural lan-guage is the basis of realizing automatic solution of mathematical problems.At present,information extraction technology is widely used in natural language field to transform unstructured knowledge into structured knowledge representation,and entity relationship extraction task is the core of information extraction technology.In this paper,we first discuss all kinds of relation extraction technology in natural lan-guage processing field,and analyze its technical characteristics.Then,on the basis of an-alyzing and summarizing the existing relation extraction model,combined with quantum mechanics theory,we propose a relation extraction model qre which combines complex words embedding and position coding.Based on the quantum superposition theory,the model uses complex words to embed and represent input sentences,uses location coding and entity context information to represent entity information,and uses projection matrix to measure the density of semantics.By comparing the qre model with the relation extrac-tion model using CNN and Gru units on semeval2010 task 8 data set,it is proved that qre model has more advantages in natural language tasks where the total vocabulary is low.Then,according to the characteristics of the elementary mathematics knowledge system and the elementary mathematics natural language such as simplicity and accuracy,a pred-icate logic representation suitable for the characteristics of the elementary mathematics natural language is constructed.Furthermore,in view of the high cost of pattern maintenance in pattern matching based relation extraction,this paper combines the proposed quantum neural network model with the traditional pattern matching based relation extraction algorithm,and proposes a hierarchical relation extraction method based on the number of sentence entities,which is implemented in the primary mathematical entity relation automatic extraction task.Through this method,five kinds of elementary mathematical problem types,plane ge-ometry,solid geometry,function,set and sequence,are tested.The recognition rate of the average problem types can reach more than 80%,which has better practical value.