arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2109.06253cs.CLcs.AIcs.LG
收藏

多语句恢复:一种减轻数据集长度偏差和束搜索性能下降的简单方法

Multi-Sentence Resampling: A Simple Approach to Alleviate Dataset Length Bias and Beam-Search Degradation

Yandex Research(Yandex研究院);Moscow Institute of Physics and Technology(莫斯科物理技术学院);HSE University(俄罗斯国立高等经济大学)

Ivan Provilkov, Andrey Malinin

摘要:

众所周知,神经机器翻译(NMT)存在束搜索问题:在某个点之后,增加波束大小会导致翻译质量的整体下降。这种影响在长句子中尤为明显。虽然已经做了很多工作来分析这种现象,主要是针对自回归NMT模型,但对其根本原因仍然没有达成共识。在这项工作中,我们分析的错误,导致主要的质量下降与大波束NMT和自动语音识别(ASR)。我们发现,一个因素,大大有助于大光束的质量下降是\textit{dataset length-bias} - \textit{NMT datasets are strongly biased towards short sentences}。为了缓解这个问题,我们提出了一种新的数据增强技术-- \textit{Multi-Sentence Resampling (MSR)}。这种技术通过连接原始数据集中的几个句子来扩展训练示例,以生成一个长的训练示例。我们证明,MSR显着减少退化与日益增长的光束大小,并提高最终的翻译质量的IWSTL $15$ En-Vi,IWSTL $17$ En-Fr,和WMT $14$ En-De数据集。

英文摘要:

Neural Machine Translation (NMT) is known to suffer from a beam-search problem: after a certain point, increasing beam size causes an overall drop in translation quality. This effect is especially pronounced for long sentences. While much work was done analyzing this phenomenon, primarily for autoregressive NMT models, there is still no consensus on its underlying cause. In this work, we analyze errors that cause major quality degradation with large beams in NMT and Automatic Speech Recognition (ASR). We show that a factor that strongly contributes to the quality degradation with large beams is \textit{dataset length-bias} - \textit{NMT datasets are strongly biased towards short sentences}. To mitigate this issue, we propose a new data augmentation technique -- \textit{Multi-Sentence Resampling (MSR)}. This technique extends the training examples by concatenating several sentences from the original dataset to make a long training example. We demonstrate that MSR significantly reduces degradation with growing beam size and improves final translation quality on the IWSTL$15$ En-Vi, IWSTL$17$ En-Fr, and WMT$14$ En-De datasets.

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。

↑