Design And Implementation Of Automatic Scoring Module Of Experimental Teaching Intelligent Management Platform

With the rise of the "artificial intelligence + education" model,using the advantages of artificial intelligence to promote the reform and innovation of education has very broad research prospects.Among them,in the field of automatic grading,automatic grading of subjective questions is the main difficulty in the current intelligent correction task for exams.The current difficulty of marking is mainly reflected in the correction of subjective questions,and the correction is completely dependent on the relevant personnel in the professional field,using natural language.Processing related technologies for automatic scoring is of great significance to promoting fairness and intelligence in education.The main research contents of this paper are as follows:1.This study conducts experiments on the automatic scoring tasks based on the text classification model and the semantic similarity model through the word vector method.In addition,CNN,LSTM,and BERT are used in the automatic scoring of these two methods.Compared with the traditional CNN and LSTM models,the BERT model based on pre-training can improve the accuracy of classification tasks by 4%-7%,and the semantic similarity task.The accuracy rate can be improved by about 3%.2.There is a method for constructing datasets of similar text pairs and dissimilar text pairs.Among them,similar text pairs can be obtained by combining answer samples with the same score to obtain positive samples,and dissimilar text pairs can be obtained by combining different question answers to obtain negative samples.This method can effectively expand the data set used for automatic scoring based on the text similarity method,and then can effectively aleviate the problem of scarcity of data samples when using the deep learning model for automatic scoring.3.In addition to the research on the automatic scoring algorithm,this paper also designs an automatic scoring system for subjective questions using Spring Boot technology.The system can be used by student users and teacher users to register and log in.The main function of the teacher user module is test question management.Through the test question management module,teachers can set questions and set the keywords,standard answers and scores of the questions.Student users log in to the system to view the test questions issued by the teacher and answer the questions.After completing the answers,they will give scores through the automatic scoring module,and then complete the automatic scoring task of online answering.4.In the automatic scoring module,this paper uses a method based on a combination of keyword features and semantic similarity to compare and score student answers and reference answers.Through the experiments in this paper,the model proposed in this paper is compared with the similarity scoring algorithm based on Jaccard.In the automatic scoring task,the model proposed in this paper has a smaller mean absolute error,which further verifies that the model in this paper has certain advantages in automatic scoring.