Machine Learning-Driven Toolkit for Sinhala Text-to-Speech and Spell Checker

This research introduces a web-based education system to assist Sri Lankan Grade 5 students overcome challenges in learning the Sinhala language. Mastering Sinhala is essential for success in the Grade 5 Scholarship Examination, where students often need help with complex grammar, syntax, and vocabulary. These challenges can significantly hinder their performance in this crucial exam. The toolkit uses advanced NLP and ML technologies to improve academic spelling and intonation during reading. Key contributions include the development of a Sinhala Spell Checker, implementing Random Forest algorithms with an 82<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">%</sup> accuracy rate, and a text-to-speech (TTS) system built on a custom neural network achieving 98% accuracy. The research employs a user-centered design approach, creating a unique dataset of over 100,000 words to train and enhance the Random Forest algorithm. Evaluation of Grade 5 students indicated that more than 90% reported improved language acquisition, marking the toolkit's effectiveness. By integrating these technologies, the project aims to address educational gaps, providing significant assistance for learners. Future research will further enhance the toolkit's accuracy and expand its application in language learning.