Improving Multi-Document Summarization through Referenced Flexible Extraction with Credit-Awareness

TL;DR

This paper introduces a hierarchical extract-then-abstract Transformer framework for multi-document summarization, utilizing loss weighting and reinforcement learning to improve extraction quality and summary coherence, achieving state-of-the-art results.

Contribution

The paper proposes a novel loss weighting mechanism and reinforcement learning approach to enhance extract-then-abstract models for multi-document summarization.

Findings

Outperforms strong baselines on multiple datasets

Achieves the best results on Multi-News, Multi-XScience, and WikiCatSum

Effectively balances training and testing objectives

Abstract

A notable challenge in Multi-Document Summarization (MDS) is the extremely-long length of the input. In this paper, we present an extract-then-abstract Transformer framework to overcome the problem. Specifically, we leverage pre-trained language models to construct a hierarchical extractor for salient sentence selection across documents and an abstractor for rewriting the selected contents as summaries. However, learning such a framework is challenging since the optimal contents for the abstractor are generally unknown. Previous works typically create pseudo extraction oracle to enable the supervised learning for both the extractor and the abstractor. Nevertheless, we argue that the performance of such methods could be restricted due to the insufficient information for prediction and inconsistent objectives between training and testing. To this end, we propose a loss weighting mechanism that makes the model aware of the unequal importance for the sentences not in the pseudo extraction oracle, and leverage the fine-tuned abstractor to generate summary references as auxiliary signals for learning the extractor. Moreover, we propose a reinforcement learning method that can efficiently apply to the extractor for harmonizing the optimization between training and testing. Experiment results show that our framework substantially outperforms strong baselines with comparable model sizes and achieves the best results on the Multi-News, Multi-XScience, and WikiCatSum corpora.