English

Distilling a Pretrained Language Model to a Multilingual ASR Model

Computation and Language 2022-06-28 v1 Artificial Intelligence Sound Audio and Speech Processing

Abstract

Multilingual speech data often suffer from long-tailed language distribution, resulting in performance degradation. However, multilingual text data is much easier to obtain, yielding a more useful general language model. Hence, we are motivated to distill the rich knowledge embedded inside a well-trained teacher text model to the student speech model. We propose a novel method called the Distilling a Language model to a Speech model (Distill-L2S), which aligns the latent representations of two different modalities. The subtle differences are handled by the shrinking mechanism, nearest-neighbor interpolation, and a learnable linear projection layer. We demonstrate the effectiveness of our distillation method by applying it to the multilingual automatic speech recognition (ASR) task. We distill the transformer-based cross-lingual language model (InfoXLM) while fine-tuning the large-scale multilingual ASR model (XLSR-wav2vec 2.0) for each language. We show the superiority of our method on 20 low-resource languages of the CommonVoice dataset with less than 100 hours of speech data.

Keywords

Cite

@article{arxiv.2206.12638,
  title  = {Distilling a Pretrained Language Model to a Multilingual ASR Model},
  author = {Kwanghee Choi and Hyung-Min Park},
  journal= {arXiv preprint arXiv:2206.12638},
  year   = {2022}
}

Comments

Accepted to Interspeech 2022. Official implementation provided in https://github.com/juice500ml/xlm_to_xlsr

R2 v1 2026-06-24T12:03:50.565Z