# On the (In)Effectiveness of Large Language Models for Chinese Text Correction

**Authors:** Yinghui Li, Haojing Huang, Shirong Ma, Yong Jiang, Yangning Li, Feng Zhou, Hai-Tao Zheng, Qingyu Zhou  
**arXiv:** [2307.09007](https://arxiv.org/abs/2307.09007)  
**Date:** 2023-07-18  
**Summary page:** https://www.summarizepaper.com/en/paper/2307.09007v1/

## Key points

- Large Language Models (LLMs) have seen significant development and progress in the Artificial Intelligence community.
- ChatGPT is a prominent representative of LLMs and has sparked extensive research on its capabilities and performance in Natural Language Processing (NLP) tasks.
- ChatGPT exhibits exceptional multilingual processing abilities, including Chinese.
- This study focuses on ChatGPT's performance in Chinese Text Correction, specifically Chinese Grammatical Error Correction (CGEC) and Chinese Spelling Check (CSC).
- The researchers find that while ChatGPT shows impressive performance in certain aspects of Chinese Text Correction, it also exhibits unsatisfactory behavior in other areas.
- These findings have implications for the application of LLMs in the Chinese NLP community.
- The research aims to facilitate the practical implementation and utilization of LLMs for text correction in Chinese.

## Summary

Recently, there has been significant development and progress in Large Language Models (LLMs), which has captivated the Artificial Intelligence community. One such model, ChatGPT, has emerged as a prominent representative of LLMs and has sparked extensive research on its capabilities and performance across various Natural Language Processing (NLP) tasks. While ChatGPT has demonstrated remarkable performance on diverse tasks, it also exhibits exceptional multilingual processing abilities, including Chinese. To delve into ChatGPT's Chinese processing capabilities, this study focuses on Chinese Text Correction—a fundamental and challenging NLP task. Specifically, the researchers evaluate ChatGPT's performance on two main scenarios of Chinese Text Correction: Chinese Grammatical Error Correction (CGEC) and Chinese Spelling Check (CSC). Through extensive analysis and comparisons with previous state-of-the-art fine-tuned models, the researchers empirically find that while ChatGPT showcases impressive performance in certain aspects of Chinese Text Correction, it also exhibits unsatisfactory behavior in other areas. These findings hold significant implications for the application of LLMs in the Chinese NLP community. By shedding light on both the strengths and limitations of ChatGPT in handling Chinese Text Correction tasks, this research aims to facilitate the practical implementation and utilization of LLMs within the context of Chinese language processing. The authors' work represents an ongoing effort to explore and enhance the effectiveness of large language models for improving text correction in Chinese.