arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 1810.03444cs.CL
收藏

基于短语的注意

Phrase-Based Attentions

Nanyang Technological University(南洋理工大学)

Phi Xuan Nguyen, Shafiq Joty

摘要:

大多数最先进的神经机器翻译系统,尽管在架构框架(例如递归,卷积)方面有所不同,但都有一个不可或缺的功能:注意力。然而,大多数现有的注意力方法是基于标记的,忽略了短语对齐的重要性,这是基于短语的统计机器翻译成功的关键因素。在本文中,我们提出了新的基于短语的注意力方法来建模的n元语法的令牌作为注意实体。我们将我们基于短语的注意力纳入最近提出的Transformer网络中,并证明我们的方法使用WMT'16训练数据在WMT newest2014上产生了1.3 BLEU的英语到德语和0.5 BLEU的德语到英语翻译任务的改进。

英文摘要:

Most state-of-the-art neural machine translation systems, despite being different in architectural skeletons (e.g. recurrence, convolutional), share an indispensable feature: the Attention. However, most existing attention methods are token-based and ignore the importance of phrasal alignments, the key ingredient for the success of phrase-based statistical machine translation. In this paper, we propose novel phrase-based attention methods to model n-grams of tokens as attention entities. We incorporate our phrase-based attentions into the recently proposed Transformer network, and demonstrate that our approach yields improvements of 1.3 BLEU for English-to-German and 0.5 BLEU for German-to-English translation tasks on WMT newstest2014 using WMT'16 training data.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑