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 1904.04079cs.CLstat.ML
收藏

神经机器翻译中参考词序灵活的可微采样

Differentiable Sampling with Flexible Reference Word Order for Neural Machine Translation

University of Maryland(马里兰大学)

Weijia Xu, Xing Niu, Marine Carpuat

摘要:

尽管在纠正机器翻译中的曝光偏差方面取得了一些经验上的成功,但预定采样算法存在一个主要缺点:它们错误地假设参考翻译和采样序列中的单词在每个时间步长都是对齐的。我们新的可微分采样算法通过优化参考可以与采样输出对齐的概率来解决这个问题,基于模型本身预测的软对齐。结果,相对于整个预测序列评估每个时间步的输出分布。实验结果表明,我们的方法提高了BLEU相比,最大似然和预定的采样基线。此外,我们的方法是简单的训练,不需要采样时间表和产量的模型,实现更大的改进与更小的光束尺寸。

英文摘要:

Despite some empirical success at correcting exposure bias in machine translation, scheduled sampling algorithms suffer from a major drawback: they incorrectly assume that words in the reference translations and in sampled sequences are aligned at each time step. Our new differentiable sampling algorithm addresses this issue by optimizing the probability that the reference can be aligned with the sampled output, based on a soft alignment predicted by the model itself. As a result, the output distribution at each time step is evaluated with respect to the whole predicted sequence. Experiments on IWSLT translation tasks show that our approach improves BLEU compared to maximum likelihood and scheduled sampling baselines. In addition, our approach is simpler to train with no need for sampling schedule and yields models that achieve larger improvements with smaller beam sizes.

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。

↑