Analysis of artificial intelligence translation combining fuzzy algorithms with improved attention mechanisms
In order to further improve the translation performance of neural machine translation system,a neural machine transla-tion model enhanced by pre-trained language model is proposed.On the one hand,the mask matrix strategy is introduced to alleviate the catastrophic forgetting problem of BERT pre-trained language model integrated into neural machine language model.On the other hand,the model can make full use of the output information of the optimized BERT and improve the performance of the model through the internal fusion and dynamic weighting of the multi-attention mechanism.The results show that when the Masking matrix coeffi-cient is 0.6,the improved Masking-BERT model with weighted fusion using the gating mechanism has the best test effect on the ex-perimental data set,and in the English-Chinese and Chinese-English translation tasks,Compared with Transformer baseline model,RNNSearch model and RNN-Deliberation model,the BLUE value is increased by 1.88 and 1.41 respectively.7.67,5.77,4.88,4.68,performance improvement is obvious.In the actual English teaching process,the AI artificial intelligence system equipped with the proposed model can not only meet the translation needs of the classroom,but also achieve high manual scores of translation accura-cy and class satisfaction,which is worthy of use and promotion in English teaching.
artificial intelligencetranslation systemimproving attention mechanismsPre-trained language modelEnglish teaching
万方数据
维普
