Virtual Speech Therapist: A Clinician-in-the-Loop AI Speech Therapy Agent for Personalized and Supervised Therapy

The paper presents the Virtual Speech Therapist (VST), an AI-driven platform designed for personalized stuttering therapy through automated workflows. VST employs deep learning for stuttering classification and utilizes multi-agent large language model (LLM) reasoning to create and refine individualized therapy plans while incorporating clinician feedback for finalization. This system aims to enhance clinical efficiency and improve outcomes for patients with speech impairments, demonstrating its potential to support evidence-based decision-making in speech therapy.