Dyslexia, a neurodevelopment disorder affecting reading and language processing, presents significant challenges in academic and professional settings. This dissertation introduces an innovative approach to generating dyslexia-friendly text using advanced natural language processing techniques. The research develops and evaluates an automated system that transforms standard text into more accessible formats for individuals with dyslexia, aiming to improve reading comprehension, speed, and overall accessibility of written materials.The methodology combines state-of-the-art language models (GPT and T5) with specialized techniques addressing dyslexia-specific challenges. A novel dataset of dyslexia-friendly text modifications was created through crowdsourcing and validated by dyslexic college students. The language models were fine-tuned on this dataset and enhanced with syllable and morphological analysis to simplify complex word structures. Two specialized dictionaries were developed: one for complex word substitution and another addressing visual confusion and phonological complexity issues common in dyslexia.A two-phase experiment with 14 dyslexic undergraduate students evaluated the system's effectiveness. Participants read original and modified versions of passages from GRE exam materials and a psychology textbook. Reading time, comprehension, and perceived difficulty were assessed using quantitative measures and qualitative feedback. The study also examined the impact of visual presentation factors such as background color, text alignment, and line spacing.Results showed improvements in reading time for modified texts, particularly after refinements based on initial findings. Readability ratings and comprehension scores yielded mixed results, while participants generally preferred the visual modifications in dyslexia-friendly versions. The study revealed individual variations in preferences and effectiveness of modifications, underscoring the need for personalized approaches.

This research contributes to accessible communication by demonstrating the potential of automated systems in generating dyslexia-friendly text. The findings have implications for educational practices, digital content creation, and assistive technology development, paving the way for more inclusive written communication.