The national strategy and the goal of cultivating comprehensive translation competence require that English translation teaching in colleges and universities should follow the wave of change and participate in the realization of the increasingly changing higher education teaching goals. In this paper, after applying the model of a multiplexed collaborative self-attention mechanism to encode and process the English utterances to be translated, the encoder-decoder attention mechanism is utilized to align the encoding-enhanced processed English sentences into clauses. The CA-CrossAttention mechanism and Softmax layer realize the automatic processing and translation process of English utterances. Subsequently, on the basis of B/S network architecture, an English translation teaching platform based on natural language processing technology is constructed by combining the automatic translation model, and an optimized teaching strategy is designed with the assistance of the platform. It was found that the student’s English language proficiency scores were significantly higher than those of the initial teaching at the 9th week of the implementation of the optimized teaching strategy (p=0.000<0.01), and the student’s overall English translation learning anxiety behavior was also alleviated to some extent. The students’ overall satisfaction with the environment perception and personal cognition of the teaching platform is high, and they are more agreeable to the evaluation and optimization suggestions of the platform’s feedback content. The purpose of this paper is to promote the reform of English translation teaching and the development of English majors, as well as guarantee the output of excellent talents in English translation.