Human-like Summarization Evaluation with ChatGPT

TL;DR

This paper investigates ChatGPT's capability to evaluate text summarization in a human-like manner, comparing it with traditional metrics and human judgment across multiple datasets.

Contribution

It demonstrates ChatGPT's effectiveness in human-like evaluation methods and its potential to outperform automatic metrics in summarization assessment.

Findings

ChatGPT performs well with Likert, pairwise, Pyramid, and factuality evaluations.

It outperforms automatic metrics on some datasets.

Prompt design influences evaluation performance.

Abstract

Evaluating text summarization is a challenging problem, and existing evaluation metrics are far from satisfactory. In this study, we explored ChatGPT's ability to perform human-like summarization evaluation using four human evaluation methods on five datasets. We found that ChatGPT was able to complete annotations relatively smoothly using Likert scale scoring, pairwise comparison, Pyramid, and binary factuality evaluation. Additionally, it outperformed commonly used automatic evaluation metrics on some datasets. Furthermore, we discussed the impact of different prompts, compared its performance with that of human evaluation, and analyzed the generated explanations and invalid responses.