Based on deep learning, this study explores a teaching model to promote college students’ English translation ability with the support of information technology. It mainly adopts the neural machine translation algorithm to make the training samples constitute data nodes with semantic feature-carrying properties through the profound training of semantic feature quantity. It then gets the best English translation utterance that recognizes the translated information. Using a combination of quantitative and qualitative methods, we explored the impact of deep learning strategies on English translation teaching in colleges and universities, as well as the evaluation of practical effects. The results of the study show that under the deep learning teaching mode, the English translation scores of college students are improved by 9.085 points, which is a significant difference (P=0.018) compared to the traditional teaching mode, indicating the excellent performance of the strategy in English translation teaching.