Multi-Reference Training with Pseudo-References for Neural Translation and Text Generation
Multi-Reference Training with Pseudo-References for Neural Translation and Text Generation
复制标题
DOI：
10.18653/v1/d18-1357
复制
发表时间：
2018-08
期刊：
影响因子：
--
通讯作者：
Renjie Zheng;Mingbo Ma;Liang Huang
中科院分区：
文献类型：
--
作者：
Renjie Zheng;Mingbo Ma;Liang Huang
神经文本生成，包括神经机器翻译、图像字幕生成和文本摘要，近来取得了相当大的成功。然而，在训练过程中，通常每个示例仅考虑一个参考译文，尽管往往存在多个参考译文，例如，在NIST机器翻译评估中有4个参考译文，在图像字幕数据中有5个参考译文。我们首先探究了在训练过程中利用多个人工参考译文的几种不同方式。但更重要的是，我们随后提出了一种算法，通过首先将现有人工参考译文压缩成格状结构，然后对其进行遍历以生成新的伪参考译文，从而生成数量呈指数级增长的伪参考译文。这些方法相较于强大的基线模型，在机器翻译（BLEU值提升1.5）和图像字幕生成（BLEU值提升3.1 / CIDEr值提升11.7）方面都带来了显著改进。
Neural text generation, including neural machine translation, image captioning, and summarization, has been quite successful recently. However, during training time, typically only one reference is considered for each example, even though there are often multiple references available, e.g., 4 references in NIST MT evaluations, and 5 references in image captioning data. We first investigate several different ways of utilizing multiple human references during training. But more importantly, we then propose an algorithm to generate exponentially many pseudo-references by first compressing existing human references into lattices and then traversing them to generate new pseudo-references. These approaches lead to substantial improvements over strong baselines in both machine translation (+1.5 BLEU) and image captioning (+3.1 BLEU / +11.7 CIDEr).