The development of science and technology makes the cooperation and cultural exchanges between various countries more and more frequent. Machine translation has the advantages of high translation efficiency and low cost, and the research devoted to machine translation has become the current trend in the cross-cultural field. Based on the application form of cross-cultural perspective transformation in English text translation, this paper proposes cross-cultural text features using the Glove model and puts forward a neuro-machine translation strategy of optimizing the Transformer model with a multi-attention mechanism, which is used to solve the task of cross-cultural English text translation. In the Chinese translation task of cross-cultural English text, the translation accuracy of the model is negatively correlated with the sequence length. The average SER is 0.551, which has good anti-confusion performance. Furthermore, it appears to have a greater lexical richness and the number of idioms used when performing the translation task. Based on this study, the proposed English text translation technique achieves better results. It can replace manual translation to accomplish complex tasks and promote the development of translation cognition.