This AI-powered web platform and mobile app combines software engineering and artificial intelligence to support Arabic-speaking children with speech sound disorders. It provides interactive therapy activities, guided sessions, and gamified progress tracking to improve pronunciation in an engaging, accurate, and child-friendly way.
At the core of the platform is a complete Mispronunciation Detection and Diagnosis (MDD) pipeline. This system leverages a custom AI model that I trained to detect mispronounced letters and identify their exact position within words, whether at the beginning (with fatha, dama, or kasra vowels after it), middle, or end.
The pipeline analyzes children’s speech at the phoneme level, provides immediate and structured feedback, and automatically generates a personalized treatment plan based on each child’s pronunciation errors.
To ensure professional integrity, the system has received clinical validation from speech therapy specialists, ensuring that its feedback and treatment recommendations are fully aligned with professional speech therapy practices.
The platform also supports the wider care ecosystem by helping therapists with advanced assessment tools, patient management portals, progress monitoring, and detailed analytics, while enabling parents to follow their child’s development and participate in supervised home practice.
In addition to the software applications, the project includes a curated dataset of 2,000 Arabic speech recordings and a fine-tuned Arabic speech model to strengthen pronunciation analysis.
It is further supported by experiments, evaluation results, diagrams, API collections, and research materials that contribute to the system’s development, validation, and continuous improvement.
