Research on Chinese-Lao Neural Machine Translation Based on Multi-Pivot
Liqing Wang, Jiaquan Li · 2023
The performance of Neural machine translation relies on a large number of the parallel corpus. However, For low-resource language, the single language pivot-based methods still difficult to meet the model training needs. To this end, multiple pivot languages are integrated into the Neural machine translation model. This thesis introduces English and Thai corpus into the Chinese-Lao Neural machine translation, the aim is to research the influence on the performance model of pivot language similarity and improve the quality of neural machine translation of Chinese-Lao. The experimental results show that the multi-pivot method could help to improve the quality of neural machine translation of low-resource language, and the influence of corpus size and quality on the results is more than that of language similarity. The multi-Pivot Multilingual Neural Machine Translation model has the highest performance advantage by sharing parameter migration across multiple pivot languages.