AgreeSum: Agreement-Oriented Multi-Document Summarization

TL;DR

This paper introduces AgreeSum, a new multi-document summarization task focused on generating summaries that reflect common and faithful information across articles, supported by a newly created dataset and strong baseline models.

Contribution

The paper proposes the AgreeSum task, creates a dedicated dataset with entailment annotations, and develops baseline models using PEGASUS and T5 to improve summary faithfulness.

Findings

Better article-summary and cluster-summary entailment in generated summaries

Supervised and unsupervised training improve summarization faithfulness

The dataset and annotations support future research in faithful abstractive summarization

Abstract

We aim to renew interest in a particular multi-document summarization (MDS) task which we call AgreeSum: agreement-oriented multi-document summarization. Given a cluster of articles, the goal is to provide abstractive summaries that represent information common and faithful to all input articles. Given the lack of existing datasets, we create a dataset for AgreeSum, and provide annotations on article-summary entailment relations for a subset of the clusters in the dataset. We aim to create strong baselines for the task by applying the top-performing pretrained single-document summarization model PEGASUS onto AgreeSum, leveraging both annotated clusters by supervised losses, and unannotated clusters by T5-based entailment-related and language-related losses. Compared to other baselines, both automatic evaluation and human evaluation show better article-summary and cluster-summary entailment in generated summaries. On a separate note, we hope that our article-summary entailment annotations contribute to the community's effort in improving abstractive summarization faithfulness.