arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 1710.03348cs.CL
收藏

神经机器翻译中的注意力关注什么?

What does Attention in Neural Machine Translation Pay Attention to?

Informatics Institute, University of Amsterdam(阿姆斯特丹大学信息学研究所)

Hamidreza Ghader, Christof Monz

摘要:

神经机器翻译中的注意力提供了在每个翻译步骤对源句子的相关部分进行编码的可能性。因此,注意力也被认为是一个对齐模型。然而,还没有专门研究注意力的工作,并提供注意力模型所学到的东西的分析。因此,问题仍然是,如何注意力是相似或不同的传统对齐。在本文中,我们提供了详细的分析注意力,并比较它与传统的对齐。我们回答的问题是,注意力是否只能模拟翻译等价物或它捕获更多的信息。我们表明,注意力是不同的对齐在某些情况下,捕捉有用的信息以外的对齐。

英文摘要:

Attention in neural machine translation provides the possibility to encode relevant parts of the source sentence at each translation step. As a result, attention is considered to be an alignment model as well. However, there is no work that specifically studies attention and provides analysis of what is being learned by attention models. Thus, the question still remains that how attention is similar or different from the traditional alignment. In this paper, we provide detailed analysis of attention and compare it to traditional alignment. We answer the question of whether attention is only capable of modelling translational equivalent or it captures more information. We show that attention is different from alignment in some cases and is capturing useful information other than alignments.

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。

↑