Research on English-Chinese Translation Quality Evaluation Agorithm Based on Cross-Language Pre-Training Mode

Abstract

To overcome the shortcomings of the quality evaluation of translation in the low-resource corpus, a cross-sentence pre-training model is proposed for English-Chinese translation. First of all, we provide an embedding technology to automatically adjust the position of words by using attention thought for reference. Then, the cross-layer language pre-training model is introduced into the reading efficiency test to solve the information sparsity caused by the low resource conditions of English. By regressing the sentence vector, the mechanical evaluation of translation quality is completed. The test results show that this model significantly improves the effectiveness of the evaluation of the quality of English-Chinese translation. Compared with the CEstmodel, the Pearson correlation coefficient of this algorithm has increased by 0.35, and the Spielbman correlation coefficient has increased by 0.15.