Unsupervised Extractive Summarization by Pre-training Hierarchical Transformers

TL;DR

This paper introduces an unsupervised extractive summarization method leveraging pre-trained hierarchical transformers and sentence-level self-attentions, achieving state-of-the-art results without labeled data.

Contribution

It proposes a novel approach using transformer self-attentions for sentence ranking in unsupervised summarization, reducing reliance on sentence positions and combining with existing models for improved performance.

Findings

Achieves state-of-the-art unsupervised summarization performance.

Less dependent on sentence positions compared to previous methods.

Combining with position-aware models yields better results.

Abstract

Unsupervised extractive document summarization aims to select important sentences from a document without using labeled summaries during training. Existing methods are mostly graph-based with sentences as nodes and edge weights measured by sentence similarities. In this work, we find that transformer attentions can be used to rank sentences for unsupervised extractive summarization. Specifically, we first pre-train a hierarchical transformer model using unlabeled documents only. Then we propose a method to rank sentences using sentence-level self-attentions and pre-training objectives. Experiments on CNN/DailyMail and New York Times datasets show our model achieves state-of-the-art performance on unsupervised summarization. We also find in experiments that our model is less dependent on sentence positions. When using a linear combination of our model and a recent unsupervised model explicitly modeling sentence positions, we obtain even better results.