PROM: A Phrase-level Copying Mechanism with Pre-training for Abstractive Summarization

TL;DR

PROM introduces a phrase-level copying mechanism with an indicator layer and auxiliary loss, significantly improving abstractive summarization performance, especially in zero-shot settings, by enhancing copying accuracy and factuality.

Contribution

The paper proposes PROM, a novel phrase-level copying mechanism with explicit n-gram attention and auxiliary loss, applicable to zero-shot and fine-tuning scenarios, advancing summarization quality.

Findings

Significant improvements on benchmark datasets.

Effective in zero-shot summarization with pre-training.

Promotes more accurate and faithful copying.

Abstract

Based on the remarkable achievements of pre-trained language models in abstractive summarization, the copying mechanism has proved helpful by improving the factuality, stability, and overall performance. This work proposes PROM, a new PhRase-level cOpying Mechanism that enhances attention on n-grams, which can be applied to zero-shot summarization with pre-training. PROM adds an indicator layer to explicitly pick up tokens in n-gram that can be copied from the source, and calculates an auxiliary loss for the copying prediction. Empirical studies show that PROM makes significant improvements in fine-tuning on benchmarks. In zero-shot setting, PROM is utilized in the self-supervised pre-training on raw corpora and provides new general baselines on a wide range of summarization datasets. Further analysis shows that PROM performs more reasonable copying and contributes to faithfulness.