arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2212.04205cs.CL
收藏

DC-MBR:用于最小贝叶斯风险解码的分布冷却

DC-MBR: Distributional Cooling for Minimum Bayesian Risk Decoding

Zhejiang University(浙江大学);Westlake University(西湖大学);Tsinghua University(清华大学);Tencent(腾讯);Westlake Institute for Advanced Study(西湖高等研究院)

Jianhao Yan, Jin Xu, Fandong Meng, Jie Zhou, Yue Zhang

AI总结:

针对MBR在标签平滑下因序列级分布过度倾斜而表现不佳的问题,提出分布冷却方法DC-MBR,通过调低Softmax温度操纵输出熵,理论证明其与预调标签平滑等价,并在NMT基准上验证了有效性。

摘要:

最小贝叶斯风险解码(MBR)已成为神经机器翻译中一种极具前景的解码算法。然而,MBR在标签平滑下表现不佳,这令人惊讶,因为标签平滑在束搜索中提供了良好的改进,并在各种任务中提高了泛化性。在这项工作中,我们表明该问题源于标签平滑在词元级和序列级分布上的不一致性。我们证明,即使标签平滑仅在词元级引起轻微变化,序列级分布也会高度倾斜。我们将此问题称为“自回归过度平滑”。为了解决这个问题,我们提出了一种简单有效的方法——分布冷却MBR(DC-MBR),该方法通过调低Softmax温度来操纵输出分布的熵。我们在理论上证明了预调标签平滑因子与分布冷却之间的等价性。在NMT基准上的大量实验验证了分布冷却在各种设置下均能改进MBR。

英文摘要:

Minimum Bayesian Risk Decoding (MBR) emerges as a promising decoding algorithm in Neural Machine Translation. However, MBR performs poorly with label smoothing, which is surprising as label smoothing provides decent improvement with beam search and improves generality in various tasks. In this work, we show that the issue arises from the un-consistency of label smoothing on the token-level and sequence-level distributions. We demonstrate that even though label smoothing only causes a slight change in the token-level, the sequence-level distribution is highly skewed. We coin the issue \emph{autoregressive over-smoothness}. To address this issue, we propose a simple and effective method, Distributional Cooling MBR (DC-MBR), which manipulates the entropy of output distributions by tuning down the Softmax temperature. We theoretically prove the equivalence between pre-tuning label smoothing factor and distributional cooling. Extensive experiments on NMT benchmarks validate that distributional cooling improves MBR in various settings.

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。

↑