Multilingual DistilWhisper: Efficient Distillation of Multi-task Speech Models via Language-Specific Experts - Télécom Paris Access content directly
Conference Papers Year : 2024

Multilingual DistilWhisper: Efficient Distillation of Multi-task Speech Models via Language-Specific Experts

Abstract

Whisper is a multitask and multilingual speech model covering 99 languages. It yields commendable automatic speech recognition (ASR) results in a subset of its covered languages, but the model still underperforms on a non-negligible number of under-represented languages, a problem exacerbated in smaller model versions. In this work, we propose DistilWhisper, an approach able to bridge the performance gap in ASR for these languages while retaining the advantages of multitask and multilingual capabilities. Our approach involves two key strategies: lightweight modular ASR fine-tuning of whisper-small using language-specific experts, and knowledge distillation from whisper-large-v2. This dual approach allows us to effectively boost ASR performance while keeping the robustness inherited from the multitask and multilingual pre-training. Results demonstrate that our approach is more effective than standard fine-tuning or LoRA adapters, boosting performance in the targeted languages for both in- and out-of-domain test sets, while introducing only a negligible parameter overhead at inference.
Fichier principal
Vignette du fichier
2311.01070.pdf (424.9 Ko) Télécharger le fichier
Origin : Files produced by the author(s)

Dates and versions

hal-04463381 , version 1 (16-02-2024)

Identifiers

Cite

Thomas Palmeira Ferraz, Marcely Zanon Boito, Caroline Brun, Vassilina Nikoulina. Multilingual DistilWhisper: Efficient Distillation of Multi-task Speech Models via Language-Specific Experts. ICASSP 2024 - 2024 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), IEEE, Apr 2024, Seoul (Korea), South Korea. ⟨hal-04463381⟩
96 View
79 Download

Altmetric

Share

Gmail Facebook X LinkedIn More