Construction And Application Of Corpus For Primary School Mathematics Learners Based On NLP

With the increasing influence of globalization,education has gradually changed from emphasizing general language to emphasizing subject language.Mathematical language is a concise,accurate language with strong generalization ability.It is not only widely used in humanities,natural sciences and other fields,but also the core of all subject languages.As the starting point of national education,primary school mathematics curriculum not only plays an important role in laying the foundation,but also attracts the attention of educators at home and abroad.Primary school mathematics is not only to impart the basic knowledge of mathematics,but also to cultivate the psychological quality of primary school students and the flexibility of their minds,which is of great help to the growth of students in the future.The study of primary school mathematics language not only helps to improve the teaching quality,but also deepens students' understanding and application of mathematics.In view of the lack of corpus support for solving primary school mathematical language problems by using intelligent technology,this paper takes the real exam questions and related knowledge points from 2013 to 2019 as the research object of corpus,and USES the MATTER cycle method to construct primary school mathematical corpus.Firstly,a model and specification are created for the specific language phenomenon,and the knowledge points and review exercises are marked according to the specification.Then,the annotated corpus just created is used for machine learning,and the results are evaluated and the model and algorithm are modified.With the modeling annotation loop and the training test loop,once the original model is added or modified,the MATTER loop will be repeated.Although this process is tedious and time-consuming,it can greatly improve the performance of the algorithm and the accuracy of the data,which provides a methodology for creating the gold standard corpus.The main research contents of this paper include the following three parts:(1)Phenomenon modeling and labeling of elementary school mathematics examination questions.This paper collected 1480 elementary school mathematicsquestions,and these questions can be divided into three categories: the number and algebra class topic,space and graphics class topic and statistics and probability class topic,according to the analysis of the knowledge system of knowledge structure,degree of difficulty of knowledge points and review the comprehensive analysis of proportion,annotation model is set up according to the model,using the GATE opposite corpus annotation tool for labeling.(2)Realization and description of automatic annotation.The semi-supervised learning is carried out based on the newly created corpus,which is divided into three parts: training set,development-test set and test set.The training set is used to train the algorithm used in the task,the development-test set is used for error analysis,and finally runs on the reserved corpus test set.Change the model based on the test results to improve the subsequent data closer to the gold standard,thereby improving the performance of the auto-tagging algorithm.(3)application of corpus for primary school mathematics learners.Based on Web technology,the corpus system of primary school mathematics is constructed,which mainly provides the function of querying knowledge points and searching related questions.The foreground interface mainly provides the fuzzy input of knowledge points and related question types.The system can quickly process the input and show the matching knowledge points and question lists to the users.Background interface is mainly for the administrator,it provides the administrator view,input,delete,modify and other management functions.By building a corpus,and according to the definition of knowledge points,inductive problem solving questions questions of method and combining with related formula rules,these rules will be marked by the Python language and storage,and use these rules to "save","general questions","problem solving" the solving process of machine,and illustrate the Python effect in solving problems,so as to build the corpus has practicability and validity.