Raphaël Bagat

BEST-RQ-BASED SELF-SUPERVISED LEARNING FOR WHISPER DOMAIN ADAPTATION

Automatic Speech Recognition (ASR) systems, despite large multilingual training, struggle in low-resource scenarios where labeled data is scarce. We propose BEARD (BEST-RQ Encoder Adaptation with Re-training and Distillation), a novel framework designed to adapt Whisper's encoder with unlabeled data…