Multilingual DistilWhisper: Efficient Distillation of Multi-task Speech Models via Language-Specific Experts
arXiv:2311.01070 · doi:10.1109/ICASSP48485.2024.10447520
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.
Accepted and Presented at IEEE ICASSP 2024 (this is extended version)
References in corpus (9)
- WavLM: Large-Scale Self-Supervised Pre-Training for Full Stack Speech Processing
- Massively Multilingual Neural Machine Translation in the Wild: Findings and Challenges
- Scaling Speech Technology to 1,000+ Languages
- Google USM: Scaling Automatic Speech Recognition Beyond 100 Languages
- PEFT-SER: On the Use of Parameter Efficient Transfer Learning Approaches For Speech Emotion Recognition Using Pre-trained Speech Models
- A Parameter-Efficient Learning Approach to Arabic Dialect Identification with Pre-Trained General-Purpose Speech Model
- ESB: A Benchmark For Multi-Domain End-to-End Speech Recognition
- Aligning Language Models with Preferences through f-divergence Minimization
- Language-Universal Adapter Learning with Knowledge Distillation for End-to-End Multilingual Speech Recognition