Digital Harf delivers AI-powered Arabic speech therapy

A team of researchers from the University of Rochester, King Faisal Specialist Hospital, and other institutions has developed Digital Harf, a clinically integrated multimodal AI system designed to extend speech and language therapy into homes for Arabic-speaking children with Autism Spectrum Disorder. The platform combines three therapeutic modules—language therapy, speech intelligibility assessment, and picture description—into a unified workflow that adapts to each child’s performance over time.

At the core of Digital Harf is the Agentic Synthetic Data Engine (ASDE), which automatically generates culturally relevant therapeutic content, including images, prompts, and language tasks. The engine operates under explicit therapeutic and cultural criteria, bypassing the scarcity of Arabic therapy materials. In expert evaluations involving 13 licensed Speech-Language Pathologists (SLPs), ASDE-generated content achieved a 90.1% clinical acceptance rate without any manual curation. SLPs also rated the platform highly for cultural and linguistic alignment, demonstrating that AI-driven therapeutic systems can be built from the ground up for underrepresented linguistic contexts.