基于多语言联合训练的汉-英-缅神经机器翻译方法(Chinese-English-Burmese Neural Machine Translation Method Based on Multilingual Joint Training)

Zhibo Man (满志博), Cunli Mao (毛存礼), Zhengtao Yu (余正涛), Xunyu Li (李训宇), Shengxiang Gao (高盛祥), Junguo Zhu (朱俊国)


Abstract
多语言神经机器翻译是解决低资源神经机器翻译的有效方法,现有方法通常依靠共享词表的方式解决英语、法语以及德语相似语言之间的多语言翻译问题。缅甸语属于一种典型的低资源语言,汉语、英语以及缅甸语之间的语言结构差异性较大,为了缓解由于差异性引起的共享词表大小受限制的问题,提出一种基于多语言联合训练的汉英缅神经机器翻译方法。在Transformer框架下将丰富的汉英平行语料与汉缅、英缅的语料进行联合训练,模型训练过程中分别在编码端和解码端将汉英缅映射在同一语义空间降低汉英缅语言结构差异性对共享词表的影响,通过共享汉英语料训练参数来弥补汉缅数据缺失的问题。实验表明在一对多、多对多的翻译场景下,提出方法相比基线模型的汉-英、英-缅以及汉-缅的BLEU值有明显的提升。
Anthology ID:
2020.ccl-1.41
Volume:
Proceedings of the 19th Chinese National Conference on Computational Linguistics
Month:
October
Year:
2020
Address:
Haikou, China
Editors:
Maosong Sun (孙茂松), Sujian Li (李素建), Yue Zhang (张岳), Yang Liu (刘洋)
Venue:
CCL
SIG:
Publisher:
Chinese Information Processing Society of China
Note:
Pages:
446–456
Language:
Chinese
URL:
https://aclanthology.org/2020.ccl-1.41
DOI:
Bibkey:
Cite (ACL):
Zhibo Man, Cunli Mao, Zhengtao Yu, Xunyu Li, Shengxiang Gao, and Junguo Zhu. 2020. 基于多语言联合训练的汉-英-缅神经机器翻译方法(Chinese-English-Burmese Neural Machine Translation Method Based on Multilingual Joint Training). In Proceedings of the 19th Chinese National Conference on Computational Linguistics, pages 446–456, Haikou, China. Chinese Information Processing Society of China.
Cite (Informal):
基于多语言联合训练的汉-英-缅神经机器翻译方法(Chinese-English-Burmese Neural Machine Translation Method Based on Multilingual Joint Training) (Man et al., CCL 2020)
Copy Citation:
PDF:
https://aclanthology.org/2020.ccl-1.41.pdf