With the emergence of generative artificial intelligence such as GPT, it provides an important tool to support the automatic study of English. This paper mainly proposes an automatic scoring method for college English based on generative AI-ChatGPT fine-tuning model. Through ChatGPT’s powerful language understanding and generative capabilities, intelligent automatic grading of students’ answers is carried out by scoring logic and algorithms. Comparative experiments are conducted on different datasets, and the conclusions show that the loss function curve of the automatic scoring algorithm proposed in this paper reaches the optimum at about 30 iterations, and its convergence speed is faster and the model performance is better. Compared with LSTM, GRU and other control algorithms, the accuracy of the method proposed in this paper reaches 89.62%, which is higher and the average absolute error is smaller. The method in this paper achieved an average score of 0.7964 on the average QWK scores of the eight topics. The attitudes of the students of the experimental and control versions towards automated scoring have mutual in difference with Sig value of 0.001.
