Towards Enhancing Coherence in Extractive Summarization: Dataset and Experiments with LLMs

TL;DR

This paper introduces a new human-annotated dataset incorporating user feedback to improve the coherence of extractive summaries generated by large language models, demonstrating significant performance gains.

Contribution

It presents a novel dataset with human-annotated coherence and user intent, and fine-tunes LLMs to enhance extractive summary coherence using this data.

Findings

Significant (~10%) Rouge-L improvement in coherence.

Fine-tuning LLMs with human feedback enhances summary quality.

Benchmarking with instruction-tuned models reveals key insights.

Abstract

Extractive summarization plays a pivotal role in natural language processing due to its wide-range applications in summarizing diverse content efficiently, while also being faithful to the original content. Despite significant advancement achieved in extractive summarization by Large Language Models (LLMs), these summaries frequently exhibit incoherence. An important aspect of the coherent summary is its readability for intended users. Although there have been many datasets and benchmarks proposed for creating coherent extractive summaries, none of them currently incorporate user intent to improve coherence in extractive summarization. Motivated by this, we propose a systematically created human-annotated dataset consisting of coherent summaries for five publicly available datasets and natural language user feedback, offering valuable insights into how to improve coherence in extractive summaries. We utilize this dataset for aligning LLMs through supervised fine-tuning with natural language human feedback to enhance the coherence of their generated summaries. Preliminary experiments with Falcon-40B and Llama-2-13B show significant performance improvements (~10% Rouge-L) in terms of producing coherent summaries. We further utilize human feedback to benchmark results over instruction-tuned models such as FLAN-T5 which resulted in several interesting findings. Data and source code are available at https://github.com/Mihir3009/Extract-AI.