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 1910.08684cs.CL
收藏

坚持事实:自信解码,实现忠实的数据到文本生成

Sticking to the Facts: Confident Decoding for Faithful Data-to-Text Generation

Google Research(谷歌研究院)

Ran Tian, Shashi Narayan, Thibault Sellam, Ankur P. Parikh

摘要:

我们解决了数据到文本生成中的幻觉问题,即,减少了源不支持的文本的生成。我们推测,幻觉可能是由编码器-解码器模型生成内容短语而不关注源引起的;因此,我们提出了一个置信度分数,以确保模型在必要时关注源,以及一个变分贝叶斯训练框架,可以从数据中学习分数。WikiBio上的实验(Lebretet等人,2016)数据集表明,根据PANITOS评分(Dhingra等人,2019年)和人类评估。我们还报告了WebNLG的强有力的结果(Gardent等人,2017)dataset.

英文摘要:

We address the issue of hallucination in data-to-text generation, i.e., reducing the generation of text that is unsupported by the source. We conjecture that hallucination can be caused by an encoder-decoder model generating content phrases without attending to the source; so we propose a confidence score to ensure that the model attends to the source whenever necessary, as well as a variational Bayes training framework that can learn the score from data. Experiments on the WikiBio (Lebretet al., 2016) dataset show that our approach is more faithful to the source than existing state-of-the-art approaches, according to both PARENT score (Dhingra et al., 2019) and human evaluation. We also report strong results on the WebNLG (Gardent et al., 2017) dataset.

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。

↑