[4G3-GS-6] Language media processing:

[4G3-GS-6-04] Proposal for a Proofreading System to Support Creators Based on Model Merging of LLMs

〇YUYA OKAMOTO¹, Naoki Mori² (1. Osaka Prefecture University, 2. Osaka Metropolitan University)

Keywords:Generative AI, Large Language Models, Proofreading Support, Writing Support

Ensuring textual correctness is vital in creative writing, necessitating thorough proofreading. However, this task reduces creators’ time for writing, prompting automated tools to ease the burden. Yet many works employ unique stylistic rules, limiting generic approaches. We propose a system supporting custom rules. We fine-tuned two RoBERTa-based models on separate datasets and merged them to build an error detection model, yielding promising results.