The Exploration of Knowledge-Preserving Prompts for Document Summarisation

TL;DR

This paper investigates using trainable prefix prompts combined with natural language prompts to improve factual consistency in document summarisation, showing that explicit factual knowledge integration enhances summary quality.

Contribution

It introduces a novel prefix-tuning approach that effectively incorporates factual knowledge into summarisation models, improving factual accuracy and overall performance.

Findings

Trainable prefixes help extract information from prompts accurately.

Knowledge-preserving summaries are more factually consistent.

Explicit factual knowledge boosts summarisation performance.

Abstract

Despite the great development of document summarisation techniques nowadays, factual inconsistencies between the generated summaries and the original texts still occur from time to time. This study explores the possibility of adopting prompts to incorporate factual knowledge into generated summaries. We specifically study prefix-tuning that uses a set of trainable continuous prefix prompts together with discrete natural language prompts to aid summary generation. Experimental results demonstrate that the trainable prefixes can help the summarisation model extract information from discrete prompts precisely, thus generating knowledge-preserving summaries that are factually consistent with the discrete prompts. The ROUGE improvements of the generated summaries indicate that explicitly adding factual knowledge into the summarisation process could boost the overall performance, showing great potential for applying it to other natural language processing tasks.