Research On The Key Technologies Of Scoring Subjective Questions In Primary School Chinese

With the rapid development and popularization of education informatization,there are more and more online examination systems and automatic scoring systems for examination papers.However,due to the complexity of Chinese grammar and the diversity of rhetoric,the automatic scoring of Chinese subjective questions is still a big challenge.In the existing research,the scoring method based on deep learning requires too many samples,and the interpretability of the scoring results is low.so it is unable to give clear evaluation information on sentence components and rhetorical scoring standards to help primary school students find mistakes and correct them.In addition,the current scoring method is only a simple accumulation of the scores of each sub item,which is difficult to describe the students' mastery of different knowledge points,and the difficulty of the test is also difficult to be reflected,which cannot scientifically reflect the ups and downs of the same examinee's multiple test scores.This paper reviews the research status of the automatic scoring technology of primary school Chinese test paper,especially the challenges and research status of the automatic scoring of primary school Chinese subjective questions.Taking the sentence type of primary school Chinese test paper as an example,we have done the following work.(1)This paper studies the scoring of subjective Chinese questions in primary schools.To meet the special needs of grammar analysis and the interpretability of scoring results,a set of subjective question scoring framework is put forward.Based on word segmentation and part-of-speech tagging,it identifies the wrong words and gives the correct answers.Based on named entity recognition,students' answers are judged to be consistent with the specified topic;Grammar is graded based on sentence component analysis and Chinese dependency parsing;and rhetorical is graded based on rules.It focuses on the syntactic analysis,semantic role labeling,and the recognition and analysis of figurative figures of speech based on LTP natural language processing package.(2)In order to make the scores of the test questions better describe the degree of the candidates' mastery of the knowledge involved in the test questions,and to some extent reflect the difference between the students' individual development and the students' overall development level,this paper gives a method to calculate the difficulty coefficient of the test questions,and studies the influence of different individual level differences on the coefficient,Finally,with the help of equal sample segmentation method,the final score of each student is corrected.The revised final score can more accurately reflect the students' mastery of the knowledge,and the discrimination of the test paper is also higher.(3)Based on Spring Boot and Layui front-end framework,a prototype system for automatic scoring of primary school Chinese subjective questions is implemented.The prototype system realizes the online examination and scoring of primary school Chinese,and supports the following types of questions: fill in the blanks,multiple-choice questions,sentence making questions and so on.The prototype system uses the subjective question scoring framework proposed in this paper,and uses the score correction algorithm proposed in this paper.The stable operation of the prototype system shows that the research is feasible.