This study focuses on the optimization and application of oral English evaluation system based on artificial intelligence (AI), and improves the accuracy and practicability of the system through technical improvement, so as to better serve oral English teaching. Firstly, the limitations of traditional manual evaluation and the shortcomings of existing AI evaluation system are analyzed, and then the system optimization scheme is put forward, including improving speech recognition algorithm, optimizing natural language processing (NLP) technology and designing personalized feedback mechanism. The system adopts multi-modal fusion architecture, combining BiLSTM-CTC speech recognition model and BERT-based grammar detection model, which significantly improves the accuracy of speech recognition and the ability of grammar error detection. In the experimental part, the English majors in a university in Xinjiang are taken as the sbjects, the learning effects of the AI evaluation system group and the traditional classroom group are compared. The results show that the AI system is significantly superior to the traditional teaching mode in pronunciation accuracy, grammar accuracy and fluency improvement, and also has significant advantages in classroom participation and learning satisfaction.
Chen et al. (Sun,) studied this question.