Evaluation in English Education Based on Computer Multimodal Discourse Analysis System

This paper proposes an English assessment system based on computer multimodal discourse analysis, which integrates multimodal data such as speech, text and facial expressions, and combines SVM and CART algorithms for automatic assessment. The system aims to improve the accuracy and comprehensiveness of English oral assessment. This paper designs a multimodal data fusion framework and applies it to the assessment of students' oral ability. The experiment uses an English oral dataset containing 300 students, and evaluates students' oral fluency, pronunciation accuracy and grammatical correctness through speech recognition, grammatical analysis and facial expression emotion analysis. The experimental results show that the multimodal analysis model based on SVM combined with decision tree has an assessment accuracy of <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathbf{9 1. 8 \%}$</tex>, which is significantly improved compared with the traditional support vector machine algorithm (84.5%) and decision tree algorithm (86.3 %). Specifically, in speech fluency assessment, the accuracy of the SVM combined with decision tree model is <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathbf{9 4. 2 \%}$</tex>, and in grammatical error detection, the accuracy is <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathbf{8 9. 6 \%}$</tex>.