Automatic Data Augmentation for Domain Adapted Fine-Tuning of Self-Supervised Speech Representations - Archive ouverte HAL Accéder directement au contenu
Communication Dans Un Congrès Année : 2023

Automatic Data Augmentation for Domain Adapted Fine-Tuning of Self-Supervised Speech Representations

Résumé

Self-Supervised Learning (SSL) has allowed leveraging large amounts of unlabeled speech data to improve the performance of speech recognition models even with small annotated datasets. Despite this, speech SSL representations may fail while facing an acoustic mismatch between the pretraining and target datasets. To address this issue, we propose a novel supervised domain adaptation method, designed for cases exhibiting such a mismatch in acoustic domains. It consists in applying properly calibrated data augmentations on a large clean dataset, bringing it closer to the target domain, and using it as part of an initial fine-tuning stage. Augmentations are automatically selected through the minimization of a conditional-dependence estimator, based on the target dataset. The approach is validated during an oracle experiment with controlled distortions and on two amateur-collected low-resource domains, reaching better performances compared to the baselines in both cases.
Fichier principal
Vignette du fichier
zaiem23_interspeech.pdf (357.04 Ko) Télécharger le fichier
Origine : Fichiers produits par l'(les) auteur(s)

Dates et versions

hal-04216177 , version 1 (23-09-2023)

Identifiants

Citer

Salah Zaiem, Titouan Parcollet, Slim Essid. Automatic Data Augmentation for Domain Adapted Fine-Tuning of Self-Supervised Speech Representations. INTERSPEECH 2023, Aug 2023, Dublin (Ireland), Ireland. pp.67-71, ⟨10.21437/Interspeech.2023-1040⟩. ⟨hal-04216177⟩
17 Consultations
37 Téléchargements

Altmetric

Partager

Gmail Facebook X LinkedIn More