mlsanityData to 7 Oct 2026

XLST: Cross-lingual Self-training to Learn Multilingual Representation for Low Resource Speech Recognition

arXiv 2103.08207 ↗First posted 15 Mar 2021eess.AS cs.CL cs.SDcited by 11

Not on the map. Why?

Abstract (arXiv)

In this paper, we propose a weakly supervised multilingual representation learning framework, called cross-lingual self-training (XLST). XLST is able to utilize a small amount of annotated data from high-resource languages to improve the representation learning on multilingual un-annotated data. Specifically, XLST uses a supervised trained model to produce initial representations and another model to learn from them, by maximizing the similarity between output embeddings of these two models. Furthermore, the moving average mechanism and multi-view data augmentation are employed, which are experimentally shown to be crucial to XLST. Comprehensive experiments have been conducted on the CommonVoice corpus to evaluate the effectiveness of XLST. Results on 5 downstream low-resource ASR tasks shows that our multilingual pretrained model achieves relatively 18.6% PER reduction over the state-of-the-art self-supervised method, with leveraging additional 100 hours of annotated English data.

Key facts. Cited by 11 papers in the map (data through 7 Oct 2026). Builds on 29 of its 45 references that are in the map. First posted on arXiv 15 Mar 2021.

Builds on

29 of 45 references.

References · 29

  1. 2020A Simple Framework for Contrastive Learning of Visual Representationscited by 11,979
  2. 2018Representation Learning with Contrastive Predictive Codingcited by 7,662
  3. 2017mixup: Beyond Empirical Risk Minimizationcited by 5,155
  4. 2020Bootstrap your own latent: A new approach to self-supervised Learningcited by 4,712
  5. 2020wav2vec 2.0: A Framework for Self-Supervised Learning of Speech Representationscited by 3,853
  6. 2020Exploring Simple Siamese Representation Learningcited by 2,853
  7. 2019Common Voice: A Massively-Multilingual Speech Corpuscited by 1,358
  8. 2019SpecAugment: A Simple Data Augmentation Method for Automatic Speech Recognitioncited by 1,355
  9. 2020Unsupervised Cross-lingual Representation Learning for Speech Recognitioncited by 496
  10. 2019vq-wav2vec: Self-Supervised Learning of Discrete Speech Representationscited by 335
  11. 2019Interpolation Consistency Training for Semi-Supervised Learningcited by 248
  12. 2019Mockingjay: Unsupervised Speech Representation Learning with Deep Bidirectional Transformer Encoderscited by 196
17 more
  1. 2020TERA: Self-Supervised Learning of Transformer Encoder Representation for Speechcited by 186
  2. 2019An Unsupervised Autoregressive Model for Speech Representation Learningcited by 185
  3. 2020Improved Noisy Student Training for Automatic Speech Recognitioncited by 139
  4. 2019Self-Training for End-to-End Speech Recognitioncited by 128
  5. 2020Unsupervised pretraining transfers well across languagescited by 100
  6. 2019Unsupervised speech representation learning using WaveNet autoencoderscited by 90
  7. 2019Transformer-based Acoustic Modeling for Hybrid Speech Recognitioncited by 86
  8. 2020DeCoAR 2.0: Deep Contextualized Acoustic Representations with Vector Quantizationcited by 77
  9. 2021UniSpeech: Unified Speech Representation Learning with Labeled and Unlabeled Datacited by 76
  10. 2020On the Comparison of Popular End-to-End Models for Large Scale Speech Recognitioncited by 62
  11. 2019Transformers with convolutional context for ASRcited by 44
  12. 2018Multilingual sequence-to-sequence speech recognition: architecture, transfer learning, and language modelingcited by 34
  13. 2017Large-Scale Domain Adaptation via Teacher-Student Learningcited by 28
  14. 2020Domain Adaptation via Teacher-Student Learning for End-to-End Speech Recognitioncited by 21
  15. 2018Sequence-based Multi-lingual Low Resource Speech Recognitioncited by 20
  16. 2015Very Deep Multilingual Convolutional Neural Networks for LVCSRcited by 10
  17. 2018Analysis of Multilingual Sequence-to-Sequence speech recognition systemscited by 2

Built on by

Cited by year

Citing papers per year: 2021: 1; 2022: 3; 2023: 7.

11 papers.

  1. 2022How Does Pre-trained Wav2Vec 2.0 Perform on Domain Shifted ASR? An Extensive Benchmark on Air Traffic Control Communicationscited by 30
  2. 2022A Survey of Multilingual Models for Automatic Speech Recognitioncited by 21
  3. 2021Wav2vec-S: Semi-Supervised Pre-Training for Low-Resource ASRcited by 7
  4. 2023Improving Speech Translation by Cross-Modal Multi-Grained Contrastive Learningcited by 7
  5. 2023A Novel Self-training Approach for Low-resource Speech Recognitioncited by 7
  6. 2022Efficient Utilization of Large Pre-Trained Models for Low Resource ASRcited by 4
  7. 2023A Virtual Simulation-Pilot Agent for Training of Air Traffic Controllerscited by 3
  8. 2023Language-Universal Phonetic Representation in Multilingual Speech Pretraining for Low-Resource Speech Recognitioncited by 3
  9. 2023Language-universal phonetic encoder for low-resource speech recognitioncited by 2
  10. 2023Unsupervised Pre-Training for Vietnamese Automatic Speech Recognition in the HYKIST Projectcited by 2
  11. 2023Unsupervised ASR via Cross-Lingual Pseudo-Labelingcited by 0