[4I2-GS-11] AI and Society:

[4I2-GS-11-05] Challenges and Potential of Applying Language Models to Evaluating the Difficulty of Children's Literature

〇Yusuke Mori¹ (1. Keio Sci-Fi Center)

Keywords:Natural Language Processing, Creative Support, Storytelling

In this study, the author considers the application of language models, particularly Large Language Models (LLMs), to evaluate the difficulty of children's literature. Referring to the works that are used as teaching materials and their grade level, the information is helpful as an objective indicator. To clarify the issues and possibilities of applying language models to such evaluation, the author attempted to evaluate the difficulty of children's literature, focusing on "Gon, the Little Fox," which is used as teaching material for fourth-grade elementary school students in Japan. It was possible to automatically determine that the work is suitable as educational material for fourth-grade elementary school students, but it was also confirmed that there were issues, such as the possibility that related information contained in the learning data could affect the judgment and that the judgment results were biased towards a specific grade in the settings of this trial.