mlsanityData to 7 Oct 2026

Learning ASR pathways: A sparse multilingual ASR model

arXiv 2209.05735 ↗First posted 13 Sep 2022eess.AS cs.CLcited by 12

Not on the map. Why?

Abstract (arXiv)

Neural network pruning compresses automatic speech recognition (ASR) models effectively. However, in multilingual ASR, language-agnostic pruning may lead to severe performance drops on some languages because language-agnostic pruning masks may not fit all languages and discard important language-specific parameters. In this work, we present ASR pathways, a sparse multilingual ASR model that activates language-specific sub-networks ("pathways"), such that the parameters for each language are learned explicitly. With the overlapping sub-networks, the shared parameters can also enable knowledge transfer for lower-resource languages via joint multilingual training. We propose a novel algorithm to learn ASR pathways, and evaluate the proposed method on 4 languages with a streaming RNN-T model. Our proposed ASR pathways outperform both dense models and a language-agnostically pruned model, and provide better performance on low-resource languages compared to the monolingual sparse models.

Key facts. Cited by 12 papers in the map (data through 7 Oct 2026). Builds on 30 of its 36 references that are in the map. First posted on arXiv 13 Sep 2022.

Builds on

30 of 36 references.

References · 30

  1. 2018The Lottery Ticket Hypothesis: Finding Sparse, Trainable Neural Networkscited by 2,250
  2. 2020Conformer: Convolution-augmented Transformer for Speech Recognitioncited by 1,838
  3. 2019SpecAugment: A Simple Data Augmentation Method for Automatic Speech Recognitioncited by 1,355
  4. 2017To prune, or not to prune: exploring the efficacy of pruning for model compressioncited by 596
  5. 2021XLS-R: Self-supervised Cross-lingual Speech Representation Learning at Scalecited by 589
  6. 2020Unsupervised Cross-lingual Representation Learning for Speech Recognitioncited by 496
  7. 2020MLS: A Large-Scale Multilingual Dataset for Speech Researchcited by 467
  8. 2020The Lottery Ticket Hypothesis for Pre-trained BERT Networkscited by 275
  9. 2018Streaming End-to-end Speech Recognition For Mobile Devicescited by 230
  10. 2020Emformer: Efficient Memory Transformer Based Acoustic Model For Low Latency Streaming Speech Recognitioncited by 111
  11. 2020Massively Multilingual ASR: 50 Languages, 1 Model, 1 Billion Parameterscited by 110
  12. 2020A Streaming On-Device End-to-End Model Surpassing Server-Side Conventional Model Quality and Latencycited by 99
18 more
  1. 2019Large-Scale Multilingual Speech Recognition with a Streaming End-to-End Modelcited by 92
  2. 2017Multilingual Speech Recognition With A Single End-To-End Modelcited by 87
  3. 2021Scaling End-to-End Models for Large-Scale Multilingual ASRcited by 61
  4. 2021Joint Unsupervised and Supervised Training for Multilingual ASRcited by 39
  5. 2017Block-Sparse Recurrent Neural Networkscited by 38
  6. 2019Optimizing Speech Recognition For The Edgecited by 32
  7. 2019Learning Sparse Sharing Architectures for Multiple Taskscited by 31
  8. 2020Efficient Knowledge Distillation for RNN-Transducer Modelscited by 29
  9. 2020Improving RNN Transducer Based ASR with Auxiliary Taskscited by 27
  10. 2022Coarsening the Granularity: Towards Structurally Sparse Lottery Ticketscited by 27
  11. 2020Dynamic Sparsity Neural Networks for Automatic Speech Recognitioncited by 23
  12. 20224-bit Conformer with Native Quantization Aware Training for Speech Recognitioncited by 22
  13. 2022LAE: Language-Aware Encoder for Monolingual and Multilingual ASRcited by 21
  14. 2021Omni-sparsity DNN: Fast Sparsity Optimization for On-Device Streaming E2E ASR via Supernetcited by 12
  15. 2022Language Adaptive Cross-lingual Speech Representation Learning with Sparse Sharing Sub-networkscited by 12
  16. 2019Improving noise robustness of automatic speech recognition via parallel data and teacher-student learningcited by 9
  17. 2021Streaming Transformer Transducer Based Speech Recognition Using Non-Causal Convolutioncited by 5
  18. 2022Learning a Dual-Mode Speech Recognition Model via Self-Pruningcited by 5

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2022: 2; 2023: 4; 2024: 3; 2025: 2; 2026: 1.

12 papers.

  1. 2024LoRA-Whisper: Parameter-Efficient and Extensible Multilingual ASRcited by 35
  2. 2023Mixture-of-Expert Conformer for Streaming Multilingual ASRcited by 17
  3. 2022Massively Multilingual ASR on 70 Languages: Tokenization, Architecture, and Generalization Capabilitiescited by 16
  4. 2023Task-Agnostic Structured Pruning of Speech Representation Modelscited by 13
  5. 2023TODM: Train Once Deploy Many Efficient Supernet-Based RNN-T Compression For On-device ASR Modelscited by 4
  6. 2023Dynamic ASR Pathways: An Adaptive Masking Approach Towards Efficient Pruning of A Multilingual ASR Modelcited by 3
  7. 2024Explaining Relationships Among Research Paperscited by 3
  8. 2022Data-Efficient Cross-Lingual Transfer with Language-Specific Subnetworkscited by 2
  9. 2024Finding Task-specific Subnetworks in Multi-task Spoken Language Understanding Modelcited by 2
  10. 2025BLR-MoE: Boosted Language-Routing Mixture of Experts for Domain-Robust Multilingual E2E ASRcited by 2
  11. 2025Effective and Efficient One-pass Compression of Speech Foundation Models Using Sparsity-aware Self-pinching Gatescited by 1
  12. 2026Improving low-resource ASR using bilingual fine-tuning with language identification: a cross-linguistic evaluationcited by 0