Two-step approaches, in which summary candidates are generated-then-reranked to return a single summary, can improve ROUGE scores over the standard single-step approach. Yet, standard decoding methods (i.e., beam search, nucleus sampling, and diverse beam search) produce candidates with redundant, and often low quality, content. In this paper, we design a novel method to generate candidates for re-ranking that addresses these issues. We ground each candidate abstract on its own unique content plan and generate distinct plan-guided abstracts using a model's top beam. More concretely, a standard language model (a BART LM) auto-regressively generates elemental discourse unit (EDU) content plans with an extractive copy mechanism. The top K beams from the content plan generator are then used to guide a separate LM, which produces a single abstractive candidate for each distinct plan. We apply an existing re-ranker (BRIO) to abstractive candidates generated from our method, as well as baseline decoding methods. We show large relevance improvements over previously published methods on widely used single document news article corpora, with ROUGE-2 F1 gains of 0.88, 2.01, and 0.38 on CNN / Dailymail, NYT, and Xsum, respectively. A human evaluation on CNN / DM validates these results. Similarly, on 1k samples from CNN / DM, we show that prompting GPT-3 to follow EDU plans outperforms sampling-based methods by 1.05 ROUGE-2 F1 points. Code to generate and realize plans is available at https://github.com/griff4692/edu-sum.

翻译：两步式方法——先生成候选摘要再通过重排序返回单个摘要——相较于标准单步方法可提升ROUGE分数。然而，标准解码方法（如束搜索、核采样及多样化束搜索）产生的候选内容存在冗余且质量偏低的问题。本文设计了一种面向重排序的新型候选生成方法以解决上述问题。我们将每篇候选摘要锚定在独特的独立内容规划上，并利用模型的最优束生成差异化的规划引导摘要。具体而言，标准语言模型（BART语言模型）通过抽取式复制机制自回归生成元素化话语单元（EDU）内容规划。内容规划生成器输出的前K个束结果，将用于引导另一个独立语言模型为每个独立规划生成单篇抽象式候选摘要。我们将现有重排序模型（BRIO）应用于本方法及基线解码方法产生的抽象式候选摘要。在广泛使用的单文档新闻语料库上，本方法较此前发表方法显示出显著相关性提升：CNN/Dailymail、NYT及Xsum数据集上ROUGE-2 F1值分别提升0.88、2.01和0.38。针对CNN/DM数据集的人工评估验证了上述结果。此外，在CNN/DM数据集1千个样本上的实验表明，引导GPT-3遵循EDU规划的方法比基于采样的方法在ROUGE-2 F1值上高出1.05分。规划生成与实现的代码已开源至https://github.com/griff4692/edu-sum。