In order to utilize the advantages of deep neural networks to further improve the interaction of English classroom teaching, this paper sets out to propose a language model based on SCN-LSTM by using CNN and LSTM as text feature extractors respectively, and jump-connecting the SCN structure in the convolutional layer to ensure the convergence speed of the language model from the existing experience of intelligent technology in English teaching. ReLU is utilized as the activation function and cross entropy as the loss function in the model training. The SCN-LSTM language model is brought into the speech recognition system as a basic model to form an intelligent English teaching system. Analyze the perplexity index of SCN-LSTM language model, bring it into the teaching scene, and analyze the language analysis effect of combining SCN-LSTM language model and speech recognition model. Design the intelligent teaching interaction mode to be brought into the English classroom teaching, and analyze the interaction degree of English teaching between the intelligent teaching mode of SCNLSTM language model and the traditional mode. The teacher-student interaction, student-student interaction, and human-computer interaction in the experimental class in the intelligent teaching mode are 7.67±2.82, 7.89±1.95, and 8.22±2.64, respectively, which are higher than that of traditional English classroom teaching.
