SCRIPTA LEARN

Dyslexia affects between 10% and 15% of school-age children, yet early detection remains a major bottleneck. Waiting times for a formal educational psychology assessment can stretch for months, while the child continues to accumulate frustration. Teachers and parents lack objective tools to identify problematic patterns in everyday writing and often have to rely on clinical judgment or intuition. As a result, many cases are detected late, once the gap with the rest of the class has already become significant.

We developed a full-stack web application that combines computer vision and natural language analysis with Claude by Anthropic to analyze children’s writing in real time. Users can upload a photo of a notebook or paste the text directly. The model identifies more than 10 types of errors — letter reversals, syllable omissions, phonetic substitutions, mirror writing, and more — classifies them by severity, detects patterns consistent with dyslexia, and generates concrete exercises adapted to each child’s specific errors. Each student has an individual profile with analysis history, a progress chart showing score evolution over time, and a record of the most frequent accumulated errors. The report can be exported as a professional PDF to bring to an educational psychologist or share with parents.

The result is a tool that compresses weeks of waiting into seconds. Teachers can assess an entire classroom objectively. Educational psychologists gain a digital companion for their sessions, with reports ready to print. Parents can act early, without waiting for an appointment, using concrete exercises to practice at home that same day. The app is live, in real use, and ready to scale through a SaaS subscription model.