This paper researches solving the culture vocabulary conversion problem for English–Chinese machine translation using the Transformer-based deep learning model. Data were drawn from multiple public English–Chinese parallel corpora, cleaned, segmented, and normalized to get high-quality training and test data. From the experimental results, it can be seen that both BLEU and TER show that the Transformer model outperforms the LSTM, GRU, and CNN models, implying an upper edge of the model in processing intricate language relations and catching contextual data. In one respect, further research identified that the deep learning model performs with high accuracy and naturalness in the process of translated cultural vocabularies, which highly supports learning and adapting the use of vocabularies in very different cultural backgrounds. Through strict experiments and data analysis, this present research shows that the machine translation system based on deep learning could reach a satisfactory realization level for translating cultural vocabulary. These results are helpful to provide implications with improved performance that are meaningful for practical applications.