Research on Dataset Generation in the Development of Large Language Models for Digital Textbooks

Recently, numerous institutions have been developing Large Language Models (LLMs). This model is ushering in revolutionary changes in various fields including society, economy, and education. The LLM in education is expanding its use, and this expansion includes providing personalized learning experiences. However, the LLM currently being developed is a general model, rather than a model specialized for a specific subject or textbook. This may have limitations in its use by teachers and learners. Therefore, in this study, the LLM development, an open-source model, is being fine-tuned using a specific dataset. Before proceeding, it is necessary to develop a specific dataset. Human-generated datasets are expensive and subject-specific, thereby having disadvantages. Therefore, in this study, we propose a method of developing a textbook dataset by applying the self-instruct technique. It is expected that a textbook-specific dataset can be developed at a low cost through this.