A Parameter-efficient Language Extension Framework for Multilingual ASR
Not on the map. Why?
Abstract (arXiv)
Covering all languages with a multilingual speech recognition model (MASR) is very difficult. Performing language extension on top of an existing MASR is a desirable choice. In this study, the MASR continual learning problem is probabilistically decomposed into language identity prediction (LP) and cross-lingual adaptation (XLA) sub-problems. Based on this, we propose an architecture-based framework for language extension that can fundamentally solve catastrophic forgetting, debudded as PELE. PELE is designed to be parameter-efficient, incrementally incorporating an add-on module to adapt to a new language. Specifically, different parameter-efficient fine-tuning (PEFT) modules and their variants are explored as potential candidates to perform XLA. Experiments are carried out on 5 new languages with a wide range of low-resourced data sizes. The best-performing PEFT candidate can achieve satisfactory performance across all languages and demonstrates superiority in three of five languages over the continual joint learning setting. Notably, PEFT methods focusing on weight parameters or input features are revealed to be limited in performance, showing significantly inferior extension capabilities compared to inserting a lightweight module in between layers such as an Adapter.
Key facts. Cited by 3 papers in the map (data through 7 Oct 2026). Builds on 27 of its 35 references that are in the map. First posted on arXiv 10 Jun 2024.
Builds on
27 of 35 references.
References · 27
- 2017Attention Is All You Needcited by 59,932
- 2014Adam: A Method for Stochastic Optimizationcited by 36,960
- 2021LoRA: Low-Rank Adaptation of Large Language Modelscited by 14,084
- 2016Overcoming catastrophic forgetting in neural networkscited by 4,718
- 2019Parameter-Efficient Transfer Learning for NLPcited by 4,195
- 2022Robust Speech Recognition via Large-Scale Weak Supervisioncited by 3,972
- 2021The Power of Scale for Parameter-Efficient Prompt Tuningcited by 3,769
- 2016Learning without Forgettingcited by 2,465
- 2020Conformer: Convolution-augmented Transformer for Speech Recognitioncited by 1,838
- 2019Common Voice: A Massively-Multilingual Speech Corpuscited by 1,358
- 2021BitFit: Simple Parameter-efficient Fine-tuning for Transformer-based Masked Language-modelscited by 1,234
- 2018Experience Replay for Continual Learningcited by 1,074
15 more
- 2018Efficient Lifelong Learning with A-GEMcited by 1,044
- 2023Google USM: Scaling Automatic Speech Recognition Beyond 100 Languagescited by 228
- 2020Supermasks in Superpositioncited by 223
- 2023Progressive Prompts: Continual Learning for Language Modelscited by 195
- 2020Continual Lifelong Learning in Natural Language Processing: A Surveycited by 152
- 2020Massively Multilingual ASR: 50 Languages, 1 Model, 1 Billion Parameterscited by 110
- 2022A Theoretical Study on Solving Continual Learningcited by 78
- 2021Scaling End-to-End Models for Large-Scale Multilingual ASRcited by 61
- 2021Recent Advances of Continual Learning in Computer Vision: An Overviewcited by 55
- 2022A Survey of Multilingual Models for Automatic Speech Recognitioncited by 21
- 2023CL-MASR: A Continual Learning Benchmark for Multilingual ASRcited by 16
- 2022Weight Averaging: A Simple Yet Effective Method to Overcome Catastrophic Forgetting in Automatic Speech Recognitioncited by 10
- 2023Master-ASR: Achieving Multilingual Scalability and Low-Resource Adaptation in ASR with Modular Learningcited by 9
- 2021Continual Learning for Monolingual End-to-End Automatic Speech Recognitioncited by 8
- 2022Losses Can Be Blessings: Routing Self-Supervised Speech Representations Towards Efficient Multilingual and Multitask Speech Processingcited by 7
Built on by
Citing papers per year: 2025: 3.
3 papers.