A Methodology and Evaluation of Easy-Read Content Generation for People with Developmental Disabilities Using Large Language Models

Easy-to-read (Plain Language) materials are essential inclusive resources for people with developmental disabilities, including those with autism spectrum disorder, and moderate to mild intellectual disabilities. However, research on Korean-language materials is still rare. This study proposed a methodology for generating easy-to-read materials using a Large Language Model (LLM). Specifically, a new Context Engineering methodology was developed using 28 quantitative elements, such as vocabulary difficulty and sentence length, and 10 qualitative elements, such as topic clarification and semantic simplification. The quality and usefulness of the generated materials were evaluated through both quantitative and qualitative assessments. The quantitative evaluation, which utilized a deep learning-based morphological analyzer, and the qualitative evaluation, which involved 19 people with developmental disabilities and 9 special education experts, both confirmed that the readability was improved compared to the original text.