APP下载

基于词形分析的汉—维机器翻译性能分析

2018-07-28古丽尼尕尔·买合木提帕力旦·吐尔逊艾斯卡尔·艾木都拉

电脑知识与技术 2018年11期

古丽尼尕尔·买合木提 帕力旦·吐尔逊 艾斯卡尔·艾木都拉

摘要:隨着我国“一带一路”战略的推进,新疆应当充分发挥其区域优势,着力于打造丝绸之路经济带核心区,其中提高汉语 - 维吾尔语(汉-维)之间机器翻译的质量有着重要的现实意义。通过对低频维吾尔语词汇进行词干词缀分割进行了基于三种不同机器翻译系统的汉-维机器翻译对比实验。该方法不仅减少词汇表大小从而减轻未登录词(Out Of Vocabulary,OOV)问题,同时也大大提升了翻译结果。其中统计机器翻译结果提升最明显,比原始提升了3.29个BLEU值。

关键词: 汉-维机器翻译; 词干词缀; 统计机器翻译; 神经机器翻译; 加强记忆翻译

中图分类号:TP391.2 文献标识码:A 文章编号:1009-3044(2018)11-0172-03

Morphological Analysis and its Effects in Chinese and Uyghur Machine Translation

Gulnigar MAHMUT,Parida TURSUN,Askar HAMDULLA

(Xinjiang University School of Information Science and Engineering, Urumqi 830046, China)

Abstract:As China's "The Belt and Road" (B&R;)strategy advances, Xinjiang should gives full play to its regional advantages and strives to create the core area of the Silk Road Economic Belt. Therefore, it is of great practical significance to improve the quality of machine translation between Chinese and Uyghur. In this paper, a comparison experiment of Chinese-Uyghur Machine Translation based on three different machine translation systems are carried out, by separating the stems and affixes of low-frequency Uyghur vocabulary. This method not only reduces the size of the vocabulary table and alleviates the problem of Out Of Vocabulary (OOV), but also greatly enhances the translation result. Among them, statistical machine translation results are the most obvious, with 3.29 more BLEU scores than the original.

Key words:Chinese and Uyghur Machine Translation; steam affix; Statistical Machine Translation; Neural Machine Translation; Memory augmented Neural Machine Translation

自2014年,神经机器翻译(Neural Machine Translation,NMT)作为一种新的机器翻译方法受到了广泛关注。神经机器翻译的核心思想是用一个神经网络(称之为编码器,Encoder)将源语言句子编码为一个稠密向量,然后再用另一个神经网络(称之为解码器,Decoder)把该向量表示转化为目标语言句子。基于长短时记忆(Long Short Term Memory,LSTM)的循环神经网络(Recurrent Neural Network,RNN)能够有效处理长距离依赖,相比于传统的统计机器翻译(Statistical Machine Translation,SMT),有效缓解了数据稀疏问题。因此NMT的翻译性能在很多语言对上明显优于SMT。然而,NMT高度依赖大规模高质量的双语数据。目前,像英语、汉语等大语种语料库的建设已经比较成熟,而像维吾尔语这样的小语种可用语言资源较少。

随着国家“一带一路”战略的深入,针对汉-维的机器翻译也成了一个热点问题。研究汉-维间机器翻译具有重要的经济价值,社会价值和理论价值。……

登录APP查看全文