Research on English Translation Error Correction by Combining Transformer and Generative Adversarial Networks
With the development of the economy,the importance of English is constantly increas-ing.In order to correct local and long-distance grammar in English translation,a Transformer model was designed for research.In order to solve the loss assessment mismatch problem faced by the Trans-former model,an adversarial learning framework was designed.It was combined with the Transformer model and used as the generator of the adversarial learning framework.In addition,the study employed the strategy gradient method to handle discrete symbol sequences.The results showed that the maxi-mum values of the Transformer model in accuracy,recall,and F0.5 were 64.77,32.03,and 53.39,re-spectively.The maximum values of accuracy,recall,and Fo.5 for the model combining Transformer and adversarial networks are 75.28,44.37,and 65.85,respectively,which are significantly better than the comparison model.It can be seen that the Transformer designed by the research institute combined with adversarial networks has better performance and can provide technical support for local and long-dis-tance grammar correction in English translation.
万方数据
维普
