BERT-SVM: A hybrid BERT and SVM method for semantic similarity matching evaluation of paired short texts in English teaching

Abstract

In today's digital era, English teaching assessment is gradually transforming to intelligence, but there are many shortcomings in the existing assessment methods. Traditional assessment methods are not only time-consuming and laborious, but also have obvious limitations in the depth and precision of semantic understanding, making it difficult to accurately capture the semantic nuances in students' short-text responses. To address these issues, we propose an innovative hybrid model, BERT-SVM, which combines the deep semantic understanding of BERT and the efficient classification ability of SVM. At the same time, a multi-layer attention mechanism is introduced to enhance the understanding of complex relationships in text. Through the multi-layer attention mechanism at the word, phrase and sentence levels, the model is able to capture the semantic features of the text more accurately, thus realizing the efficient evaluation of the semantic similarity of short texts. The experimental results show that the BERT-SVM model achieves 90 % accuracy in the classification task, which is significantly better than the traditional methods, proving its effectiveness and reliability in the assessment of English teaching.

Keywords

• Artificial intelligence
• Assessment model
• A hybrid BERT and SVM method
• English teaching
• Instructional reform