In order to improve the efficiency, accuracy and context comprehension of translation and accurately present the source language content, deep learning-based neural machine translation (NMT) method has become the current mainstream machine translation method. In this paper, a convolutional neural network is integrated into the Transformer translation model, and an improved CNN-Transformer English translation model is constructed, and the optimisation effect of the model on semantic accuracy and contextual comprehension is verified in an English-Chinese parallel corpus with bilingual evaluation of alternatives (BLEU) and perplexity level (PPL) as evaluation metrics, respectively.The CNN- Transformer model has a training time of 17h, which is shortened by 3h and 9h compared to the CNN model and the Transformer machine translation model, respectively, indicating that the CNN-Transformer algorithm can improve the training speed and computational efficiency of the machine translation model. The BLEU values of the CNN-Transformer model are higher than those of the basic CNN model and the Transformer machine translation model in the four English slicing granularities of words, syllables, subwords and characters, and the semantic accuracy is higher.The PPL values of the CNN-Transformer model in English-Chinese bilingual translation are always smaller than those of the other two models, and the decrease of PPL values is the largest, which indicates that the model in this paper can perform English-Chinese bilingual translation with the PPL values of the other two models, and the decrease of PPL values of the other two models is the largest. Maximum, indicating that the model in this paper has a higher fluency when performing English-Chinese bilingual translation, and has the best optimisation effect on context comprehension among the three models. This paper provides a feasible path to optimise the semantic accuracy and contextual comprehension of English translation.