Identifying Dyslexia of Traditional Chinese by Using Machine Learning Methods

Dyslexia can significantly impede reading and writing skills, hindering learning progress. Early identification of dyslexic students is crucial to provide timely intervention and support but waiting times for assessment services are often long. To address this challenge, we propose a dyslexia identification system built on AWS cloud platform. The cloud-based system is accessible via a mobile application. We collected handwriting data on 27 representative traditional Chinese characters and recruited 107 children aged 5-7 in Hong Kong for the first data collection round. We developed feature extraction algorithms, trained and evaluated 6 models for each character, and summarized the results for all 27 characters. We conducted a user acceptance test and presented our findings and conclusions in this research. The experimental result shows 84.3% real-world accuracy on dyslexia diagnosis by using our cloud-based dyslexia identification system.