# Summaries as Captions: Generating Figure Captions for Scientific Documents with Automated Text Summarization

## TL;DR

This paper proposes a novel approach to generate scientific figure captions by framing it as a text summarization task, fine-tuning a pre-trained model to improve caption quality over vision-based methods.

## Contribution

It introduces a text summarization-based method for figure caption generation, outperforming prior vision-based approaches on large-scale scientific datasets.

## Key findings

- Outperforms vision-based methods in automatic and human evaluations.
- Effectively handles low-quality author-written captions.
- Addresses lack of standards for good figure captions.

## Abstract

Good figure captions help paper readers understand complex scientific figures. Unfortunately, even published papers often have poorly written captions. Automatic caption generation could aid paper writers by providing good starting captions that can be refined for better quality. Prior work often treated figure caption generation as a vision-to-language task. In this paper, we show that it can be more effectively tackled as a text summarization task in scientific documents. We fine-tuned PEGASUS, a pre-trained abstractive summarization model, to specifically summarize figure-referencing paragraphs (e.g., "Figure 3 shows...") into figure captions. Experiments on large-scale arXiv figures show that our method outperforms prior vision methods in both automatic and human evaluations. We further conducted an in-depth investigation focused on two key challenges: (i) the common presence of low-quality author-written captions and (ii) the lack of clear standards for good captions. Our code and data are available at: https://github.com/Crowd-AI-Lab/Generating-Figure-Captions-as-a-Text-Summarization-Task.