BottleSum: Unsupervised and Self-supervised Sentence Summarization using the Information Bottleneck Principle

TL;DR

This paper introduces BottleSum, an unsupervised and self-supervised sentence summarization method based on the Information Bottleneck principle, leveraging pretrained language models to produce concise summaries without direct supervision.

Contribution

It presents a novel application of the Information Bottleneck principle to unsupervised and self-supervised sentence summarization using pretrained language models.

Findings

Extractive summarization outperforms other unsupervised models on multiple metrics.

Self-supervised abstractive summaries outperform unsupervised baselines in human evaluations.

The approach is efficient and does not require labeled data.

Abstract

The principle of the Information Bottleneck (Tishby et al. 1999) is to produce a summary of information X optimized to predict some other relevant information Y. In this paper, we propose a novel approach to unsupervised sentence summarization by mapping the Information Bottleneck principle to a conditional language modelling objective: given a sentence, our approach seeks a compressed sentence that can best predict the next sentence. Our iterative algorithm under the Information Bottleneck objective searches gradually shorter subsequences of the given sentence while maximizing the probability of the next sentence conditioned on the summary. Using only pretrained language models with no direct supervision, our approach can efficiently perform extractive sentence summarization over a large corpus. Building on our unsupervised extractive summarization (BottleSumEx), we then present a new approach to self-supervised abstractive summarization (BottleSumSelf), where a transformer-based language model is trained on the output summaries of our unsupervised method. Empirical results demonstrate that our extractive method outperforms other unsupervised models on multiple automatic metrics. In addition, we find that our self-supervised abstractive model outperforms unsupervised baselines (including our own) by human evaluation along multiple attributes.