In order to improve the effect of real-time portable English translation, this paper proposes a portable machine translation method based on BERT, thereby improving the accuracy and coherence of domain translation. The semantic information of the intermediate state is extracted through BERT+SpaCy, and the word vector containing syntactic knowledge is trained. At the same time, the bilingual corpus consisting of the intermediate state and English in the network security field is used as the input of the model to improve the performance of the translation model. In addition, this paper incorporates intelligent translation literacy into the model and constructs a conceptual framework of intelligent translation literacy from four dimensions: thinking, knowledge, ability, and ethics. Finally, this paper combines experimental analysis to verify that the model meets the experimental requirements in terms of complexity and accuracy, and is in line with the actual situation of portable translation. In the future, this method can be directly integrated into the portable translation software to improve the portable translation effect.