Pivot-based Chinese-Vietnamese joint training neural machine translation method

Abstract

The invention relates to a pivot-based Chinese-Vietnamese joint training neural machine translation method, and belongs to the technical field of natural language processing. The method comprises thesteps that firstly, Chinese and Vietnamese word vector representations are obtained by training a translation model by Chinese-Vietnamese parallel corpora; secondly, English as a pivot language jointly trains Chinese-English and English-Vietnamese translation models, and calculating and optimizing vector representations of Chinese and Vietnamese of Chinese-English, English-Vietnamese translation models and vector representations of Chinese and Vietnamese acquired by a Chinese-Vietnamese model to joint train Chinese and Vietnamese. A Chinese-Vietnamese corpus, a Chinese-English corpus and an English-Vietnamese corpus are jointly trained. An English pivot corpus is fully utilized to improve the Chinese-English machine translation performance, so that the problem of poor translation model performance caused by missing of the Chinese-English parallel corpus is solved.

Description

technical field [0001] The invention relates to a pivot-based Chinese-Vietnamese joint training neural machine translation method, belonging to the technical field of natural language processing. Background technique [0002] Machine translation is an effective tool for large-scale language translation. In recent years, the exchanges and cooperation between China and Vietnam have become closer and closer, and machine translation is a more effective way of cross-language information exchange. Therefore, it is very important to study Chinese-Vietnamese machine translation. application value. [0003] Neural machine translation is a machine translation method proposed in 2014. At present, the mainstream neural machine translation models all adopt the encoder-decoder architecture. Neural machine translation has achieved good translation performance on language pairs with large-scale parallel corpora, but in low-resource scenarios, the translation quality of neural machine trans...

AI Technical Summary

Problems solved by technology

[0006] The invention provides a pivot-based Chinese-Vietnamese joint training neural machine translation method to solve the problem that the translation quality of neural machine translation is lower than that of statistical machine translation in low-resource scenarios; it solves the problem of low resources in Chinese-Vietnamese In terms of language pairing, the scale and quality of the Chinese-Vietnamese parallel corpus lead to poor performance of Chinese-Vietnamese machine translation

Images