Joint Retrieval and Generation Training for Grounded Text Generation
Joint Retrieval and Generation Training for Grounded Text Generation
复制标题
DOI：
--
复制
发表时间：
2021
期刊：
影响因子：
--
通讯作者：
Yizhe Zhang;Siqi Sun;Xiang Gao;Yuwei Fang;Chris Brockett;Michel Galley;Jianfeng Gao;Bill Dolan
中科院分区：
文献类型：
--
作者：
Yizhe Zhang;Siqi Sun;Xiang Gao;Yuwei Fang;Chris Brockett;Michel Galley;Jianfeng Gao;Bill Dolan
最近在大规模预训练方面的进展，如GPT-3，允许从给定的提示生成看似高质量的文本。然而，这样的生成系统经常受到幻觉事实问题的困扰，并且其设计本身并不包含有用的外部信息。接地生成模型似乎提供了补救措施，但它们的训练通常依赖于很少可用的并行数据，其中提供了上下文的相应信息和相关文件。我们提出了一个框架，通过在语言模型信号上联合训练接地生成器和文档检索器来缓解这种数据约束。该模型学习在生成中奖励具有最高效用的文档检索，并使用混合专家（MoE）集成仔细地将它们组合在一起以生成后续文本。我们证明了生成器和检索器都可以利用这种联合训练并协同工作，在散文和对话生成中产生更多信息和相关的文本
Recent advances in large-scale pre-training such as GPT-3 allow seemingly high quality text to be generated from a given prompt. However, such generation systems often suffer from problems of hallucinated facts, and are not inherently designed to incorporate useful external information. Grounded generation models appear to offer remedies, but their training typically relies on rarely-available parallel data where corresponding informationrelevant documents are provided for context. We propose a framework that alleviates this data constraint by jointly training a grounded generator and document retriever on the language model signal. The model learns to reward retrieval of the documents with the highest utility in generation, and attentively combines them using a Mixture-of-Experts (MoE) ensemble to generate follow-on text. We demonstrate that both generator and retriever can take advantage of this joint training and work synergistically to produce more informative and relevant text in both prose and dialogue generation.1