Research And Implement Of Automatic Scoring System Based On Semantic Analysis

With the upsurge of artificial intelligence,more and more artificial intelligence applications based on big data have been researched and solved.Text is an abstract form of information bearing.Machine understanding of text can better reflect the intelligence of artificial intelligence.Machine translation,question answering system,chatbot,and sentiment analysis task all need to handle with text.The study of text processing algorithms used in educational intelligence systems can help the intelligent development of Internet education,and users can learn knowledge more efficiently.The main research work of the paper is as follows:(1)Analyze the subjective datasets from ASAP,and then study the three types of scoring models based on discrete representation,distributed representation and decentralized representation.The comparative analysis is based on the Brown word clustering scoring method,the evaluation based on the potential Dirichlet distribution method,and the neural network-based scoring method on the data set.It also compares and analyzes the performance of random forest regression algorithm and the XGBoost regression algorithm on feature fusion.(2)Design and implement an automatic scoring system based on semantic analysis,complete system requirements,overall architecture design,functional module design and database design,use the SSH framework to implement the user login module,test questions exercise module,automatic scoring module,user center module,and administrator function module in the system.(3)Apply the semantic matching model of SimNet to the system's automatic scoring module,and evaluate the performance of the SimNet on the Chinese historical data set.verify whether the semantic matching model of SimNet can meet the application requirement of the system.The innovation of this article is mainly reflected in:(1)The text is trained on the subjective question dataset of the ASAP using a neural network-based scoring algorithm,and the score of the algorithm is higher than the score of top1 of the data set on the Kaggle website.It is proved that in the subjective question dataset of ASAP,the ability of neural network to extract features is stronger than machine learning,and the feature extraction method is also more simple.(2)The semantic matching model of SimNet is applied to the system's automatic scoring module for the score evaluation of the users' answers.This paper evaluates the performance of the semantic matching model of SimNet on Chinese historical data sets.The results show that the semantic matching model of SimNet can meet the application requirement of the system.