In order to alleviate the problems of unregistered words,over translation,and missing translations in existing neural machine translation systems,a Chinese English automatic translation system based on improved data generalization was proposed.Da-ta enhancement and decoding strategies were incorporated in the process to obtain high-quality pseudobilingual sentence pairs,effec-tively avoiding the system from saving multiple models;In addition,a Chinese English machine translation model incorporating multi-ple coverage mechanisms is introduced to alleviate the occurrence of over translation and missed translation.The results show that the research method reached a stable state during the 41st and 19th iterations;When the training data sample set is 6 × At 105,the re-search method MarcoF1 had a high value of 97.8％;In the actual effect comparison,when the source language sentence length inter-val is higher than 50,the BLEU value of the research method is as high as 98.23％.The above data indicate that the research method can effectively improve the translation accuracy of Chinese English automatic translation systems,and can translate sentences of differ-ent lengths,providing a new reference scheme for the subsequent performance improvement of machine automatic translation systems.
维普
万方数据