mlsanityData to 7 Oct 2026

MUST: A Multilingual Student-Teacher Learning approach for low-resource speech recognition

arXiv 2310.18865 ↗First posted 29 Oct 2023cs.CL cs.SD eess.AScited by 0

Not on the map. Why?

Abstract (arXiv)

Student-teacher learning or knowledge distillation (KD) has been previously used to address data scarcity issue for training of speech recognition (ASR) systems. However, a limitation of KD training is that the student model classes must be a proper or improper subset of the teacher model classes. It prevents distillation from even acoustically similar languages if the character sets are not same. In this work, the aforementioned limitation is addressed by proposing a MUltilingual Student-Teacher (MUST) learning which exploits a posteriors mapping approach. A pre-trained mapping model is used to map posteriors from a teacher language to the student language ASR. These mapped posteriors are used as soft labels for KD learning. Various teacher ensemble schemes are experimented to train an ASR model for low-resource languages. A model trained with MUST learning reduces relative character error rate (CER) up to 9.5% in comparison with a baseline monolingual ASR.

Key facts. No citing papers in the map yet (data through 7 Oct 2026). Builds on 16 of its 30 references that are in the map. First posted on arXiv 29 Oct 2023.

Builds on

16 of 30 references.

References · 16

  1. 2015Distilling the Knowledge in a Neural Networkcited by 10,076
  2. 2014FitNets: Hints for Thin Deep Netscited by 1,619
  3. 2018Subword Regularization: Improving Neural Network Translation Models with Multiple Subword Candidatescited by 632
  4. 2021SpeechBrain: A General-Purpose Speech Toolkitcited by 404
  5. 2016Joint CTC-Attention based End-to-End Speech Recognition using Multi-task Learningcited by 254
  6. 2020Massively Multilingual ASR: 50 Languages, 1 Model, 1 Billion Parameterscited by 110
  7. 2021Simple and Effective Zero-shot Cross-lingual Phoneme Recognitioncited by 79
  8. 2021Mosaicking to Distill: Knowledge Distillation from Out-of-Domain Datacited by 31
  9. 2021Embracing the Dark Knowledge: Domain Generalization Using Regularized Knowledge Distillationcited by 28
  10. 2021Deciphering Speech: a Zero-Resource Approach to Cross-Lingual Transfer in ASRcited by 11
  11. 2023Ensemble knowledge distillation of self-supervised speech modelscited by 10
  12. 2022Investigating the Impact of Cross-lingual Acoustic-Phonetic Similarities on Multilingual Speech Recognitioncited by 5
4 more
  1. 2021Domain Generalization on Efficient Acoustic Scene Classification using Residual Normalizationcited by 4
  2. 2022Non-Linear Pairwise Language Mappings for Low-Resource Multilingual Acoustic Model Fusioncited by 2
  3. 2023Towards domain generalisation in ASR with elitist sampling and ensemble knowledge distillationcited by 1
  4. 2023Learning Cross-lingual Mappings for Data Augmentation to Improve Low-Resource Speech Recognitioncited by 1

Built on by

Not cited yet.