In order to improve students’ English translation learning effect, this paper combines the idea of generative artificial intelligence to design an English translation teaching system combined with deep learning. The system talks with students through man-computer interaction, so that students can gradually improve their translation ability in the dialogue. The model built by combining deep learning and artificial intelligence in this paper has improved the quality of translation generation to a certain extent, and achieved good results in MSK long and short sentence indicators. In addition, it also has a certain emotion recognition ability, which can analyze the emotion of users’ conversations and generate corresponding conversation replies. Finally, combined with experimental analysis, this paper verifies that the performance of the generative dialogue system meets the actual teaching needs, and the evaluation of teachers and students shows that the system has a good user experience.