Harnessing Rule-Based Reinforcement Learning for Enhanced Grammatical Error Correction

TL;DR

This paper introduces a novel rule-based reinforcement learning framework for grammatical error correction, demonstrating state-of-the-art results on Chinese datasets and highlighting the benefits of RL in guiding large language models.

Contribution

It presents a new rule-based RL approach that improves grammatical error correction by enhancing recall and control over LLM outputs, surpassing traditional supervised methods.

Findings

Achieved state-of-the-art performance on Chinese GEC datasets.

Significant increase in recall compared to existing methods.

Demonstrated the effectiveness of RL in guiding LLMs for GEC.

Abstract

Grammatical error correction is a significant task in NLP. Traditional methods based on encoder-decoder models have achieved certain success, but the application of LLMs in this field is still underexplored. Current research predominantly relies on supervised fine-tuning to train LLMs to directly generate the corrected sentence, which limits the model's powerful reasoning ability. To address this limitation, we propose a novel framework based on Rule-Based RL. Through experiments on the Chinese datasets, our Rule-Based RL framework achieves \textbf{state-of-the-art }performance, with a notable increase in \textbf{recall}. This result clearly highlights the advantages of using RL to steer LLMs, offering a more controllable and reliable paradigm for future development in GEC.