Performance comparison of multiple scoring algorithms: A study of automatic scoring of English text.

Objective:In order to save teachers' correcting time, improve the accuracy and efficiency of English composition grading.Methods: This paper briefly introduces the algorithm of deep sentence smoothness and text semantic matching based on graph neural network, and then designs an automatic scoring algorithm for English text. Result: The experimental data was collected from 12,000 essays written by international students in the United States in the Pratt & Whitney Foundation's Automated Student Value Assessment Project (ASAP), and these data were graded through a comparative experiment,Through comparative tests, the automatic scoring algorithm designed in this paper can achieve better scoring results and better handle automatic essay scoring problems. Among all the experimental mean values of evaluation methods, the experimental mean value of the algorithm designed in this paper is 0.790, the smoothness algorithm is 0.768, and the text matching vector is 0.759. The experimental mean values of the other two traditional automatic scoring algorithms are 0.710 and 0.712 respectively, and the results are lower than the algorithm designed in this paper. Conclusion: According to the experimental results, it can be concluded that good feature selection can give good scoring performance to the algorithm and cope with the problem of automatic scoring. At the same time, it also confirms the feasibility of the algorithm designed in this paper, which can be effectively applied in practical English composition scoring. [ABSTRACT FROM AUTHOR]