Text Refinement Using Large Language Models

Abstract

Large language models (LLMs) are widely used for revising human-written text to make the writing process more efficient. While there have been many studies on error correction or proofreading, text refinement is understudied and still challenging. In this study, we use LLMs to revise the abstract of academic papers written in Japanese. We propose prompts for LLMs to provide detailed feedback considering the characteristics of academic writing. In our experiments, the number of comments related to academic writing-specific expressions increased, and the proportion of appropriate feedback also improved.