Audio & Speech

CLAP-Based Automatic Word Naming Recognition in Post-Stroke Aphasia

A new AI breakthrough could transform how we diagnose and treat speech disorders...

Deep Dive

Researchers have developed a new AI system that can accurately recognize words spoken by post-stroke patients with aphasia, a condition causing severe speech disfluencies. Using a Contrastive Language-Audio Pretraining (CLAP) approach, the model treats recognition as an audio-text matching problem in a shared embedding space. Tested on French patient datasets, it achieved up to 90% accuracy, significantly outperforming existing conventional speech recognition and classification-based methods that struggle with mispronunciations.

Why It Matters

This enables reliable, automated assessment of speech recovery, paving the way for more accessible and frequent patient monitoring outside clinical settings.