BEST-RQ-Based Self-Supervised Learning for Whisper Domain Adaptation

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Bagat, Raphaël, Illina, Irina, Vincent, Emmanuel
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866909034461790208
author Bagat, Raphaël
Illina, Irina
Vincent, Emmanuel
author_facet Bagat, Raphaël
Illina, Irina
Vincent, Emmanuel
contents 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. Unlike traditional self-supervised learning methods, BEARD uniquely combines a BEST-RQ objective with knowledge distillation from a frozen teacher encoder, ensuring the encoder's complementarity with the pre-trained decoder. Our experiments focus on the ATCO2 corpus from the challenging Air Traffic Control (ATC) communications domain, characterized by non-native speech, noise, and specialized phraseology. Using about 5,000 hours of untranscribed speech for BEARD and 2 hours of transcribed speech for fine-tuning, the proposed approach significantly outperforms previous baseline and fine-tuned model, achieving a relative improvement of 12% compared to the fine-tuned model. To the best of our knowledge, this is the first work to use a self-supervised learning objective for domain adaptation of Whisper.
format Preprint
id arxiv_https___arxiv_org_abs_2510_24570
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle BEST-RQ-Based Self-Supervised Learning for Whisper Domain Adaptation
Bagat, Raphaël
Illina, Irina
Vincent, Emmanuel
Computation and Language
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. Unlike traditional self-supervised learning methods, BEARD uniquely combines a BEST-RQ objective with knowledge distillation from a frozen teacher encoder, ensuring the encoder's complementarity with the pre-trained decoder. Our experiments focus on the ATCO2 corpus from the challenging Air Traffic Control (ATC) communications domain, characterized by non-native speech, noise, and specialized phraseology. Using about 5,000 hours of untranscribed speech for BEARD and 2 hours of transcribed speech for fine-tuning, the proposed approach significantly outperforms previous baseline and fine-tuned model, achieving a relative improvement of 12% compared to the fine-tuned model. To the best of our knowledge, this is the first work to use a self-supervised learning objective for domain adaptation of Whisper.
title BEST-RQ-Based Self-Supervised Learning for Whisper Domain Adaptation
topic Computation and Language
url https://arxiv.org/abs/2510.24570