A Network Security Machine Translation Model Based on Intermediate State
Aiming at the problems of Chinese-to-English machine translation in the field of network security,such as the existence of a large number of jargons,the small size of the data and the great differences in the syntactic structure of different languages,a middle-state-based machine translation model for network security was proposed.By studying the differences in the syntactic structure of Chinese and English,the rules of Chinese conversion intermediate state were defined,BERT and SpaCy were combined for generating word vectors containing semantic embeddings in intermediate state,BERT was used to capture the contextual semantic information of sentences,and the dependencies were resolved by SpaCy,and the two were integrated into a high-dimensional feature vector.The experimental results show that the proposed translation model,based on the Transformer translation model,achieves a BLEU value of 27.0 for the low-resource cybersecurity bilingual dataset,which is an improvement of 1.0,and a BLEU value of 21.3 for the WMT18 public bilingual dataset,which is an improvement of 0.7 BLEU values.It can be seen that through intermediate states as well as capturing contextual semantic information and syntactic structure information,the model is able to deal with technical terms,complex syntax and other problems more accurately.
network securitymachine translationdependency analysistransformer model
万方数据
维普
