利用大型語言模型改寫中文教育文本可讀性:策略與挑戰

Rewriting educational material has long been an important but labor-intensive task. This paper explores the potential of using open-source large language models (LLMs) to rewrite Chinese educational materials for different grade-level readability. The main methodology is zero-shot prompting approaches without any fine-tuning. While LLMs demonstrated excellent performance in rewriting English materials, their effectiveness for Chinese materials was limited. To address this, we proposed and evaluated three strategies: cross-lingual rewriting, specific grade-level targeting, and iterative rewriting. Our findings suggest that these strategies improved the rewriting of Chinese educational materials, though challenges remain. We provide an in-depth analysis of the advantages and limitations of LLMs in educational material rewriting and discuss potential educational applications.