Generating EDU Extracts for Plan-Guided Summary Re-Ranking

Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics Volume 1: Long Papers, pages 2680–2697 July 9-14, 2023 ©2023 Association for Computational Linguistics Generating EDU Extracts for Plan-Guided Summary Re-Ranking Griffin Adams♠,♣ griffin.adams@columbia.edu Alexander R. Fabbri♢ afabbri@salesforce.com Faisal Ladhak ♠ faisal@cs.columbia.edu Kathleen McKeown ♠ kathy@cs.columbia.edu Noémie Elhadad♠,♣ noemie.elhadad@columbia.edu Salesforce Research♢ Columbia University: Computer Science♠, Biomedical Informatics♣ Abstract 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. 1 Introduction Generating diverse abstracts and then re-ranking can lead to large performance gains (in ROUGE) (Liu et al., 2022b; Ravaut et al., 2022a) over the standard approach of generating a single summary. Typically, diversity is controlled for at the token-level by modifying beam search to introduce sampling (top-K (Fan et al., 2018), nucleus (Holtzman et al., 2020)) or directly penalize repetition (Vijayakumar et al., 2016). <e>(CNN)There was a street named after Chuck Norris,</e><e> but they had to change the name because nobody crosses Chuck Norris and lives.</e> <e>Chuck Norris counted to infinity.</e> <e>Twice.</e> <e>Death once had a near-Chuck Norris experience.</e> <e>Chuck Norris is celebrating his 75th birthday</e> <e> -- but the calendar is only allowed to turn 39.</e> <e>That last one is true (well, the first part, anyway).</e> <e>The actor, martial-arts star and world's favorite tough-guy joke subject was born March 10, 1940,</e><e> which makes him 75 today.</e> <e>Or perhaps he IS 39.</e> <e>Because maybe YOU can't beat time,</e><e> but Chuck Norris can beat anything.</e> <e>Happy birthday!</e> Tuesday is Chuck Norris' 75th birthday . The actor and martial arts master is now known as subject of tough-guy one-liners . Input Document Reference Summary Figure 1: EDU Plan-Guided Abstraction (PGA). EDU spans form the oracle content plan, while EDU spans …

Content truncated to 500 words