VietASR: Achieving Industry-level Vietnamese ASR with 50-hour labeled data and Large-Scale Speech Pretraining

Fuente: arXiv
Salvato in:
Dettagli Bibliografici
Autori principali: Zhuo, Jianheng, Yang, Yifan, Shao, Yiwen, Xu, Yong, Yu, Dong, Yu, Kai, Chen, Xie
Natura: Preprint
Pubblicazione: 2025
Soggetti:
Accesso online:
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866910973614358528
author Zhuo, Jianheng
Yang, Yifan
Shao, Yiwen
Xu, Yong
Yu, Dong
Yu, Kai
Chen, Xie
author_facet Zhuo, Jianheng
Yang, Yifan
Shao, Yiwen
Xu, Yong
Yu, Dong
Yu, Kai
Chen, Xie
contents Automatic speech recognition (ASR) has made remarkable progress but heavily relies on large-scale labeled data, which is scarce for low-resource languages like Vietnamese. While existing systems such as Whisper, USM, and MMS achieve promising performance, their efficacy remains inadequate in terms of training costs, latency, and accessibility. To address these issues, we propose VietASR, a novel ASR training pipeline that leverages vast amounts of unlabeled data and a small set of labeled data. Through multi-iteration ASR-biased self-supervised learning on a large-scale unlabeled dataset, VietASR offers a cost-effective and practical solution for enhancing ASR performance. Experiments demonstrate that pre-training on 70,000-hour unlabeled data and fine-tuning on merely 50-hour labeled data yield a lightweight but powerful ASR model. It outperforms Whisper Large-v3 and commercial ASR systems on real-world data. Our code and models will be open-sourced to facilitate research in low-resource ASR.
format Preprint
id arxiv_https___arxiv_org_abs_2505_21527
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle VietASR: Achieving Industry-level Vietnamese ASR with 50-hour labeled data and Large-Scale Speech Pretraining
Zhuo, Jianheng
Yang, Yifan
Shao, Yiwen
Xu, Yong
Yu, Dong
Yu, Kai
Chen, Xie
Audio and Speech Processing
Artificial Intelligence
Computation and Language
Sound
Automatic speech recognition (ASR) has made remarkable progress but heavily relies on large-scale labeled data, which is scarce for low-resource languages like Vietnamese. While existing systems such as Whisper, USM, and MMS achieve promising performance, their efficacy remains inadequate in terms of training costs, latency, and accessibility. To address these issues, we propose VietASR, a novel ASR training pipeline that leverages vast amounts of unlabeled data and a small set of labeled data. Through multi-iteration ASR-biased self-supervised learning on a large-scale unlabeled dataset, VietASR offers a cost-effective and practical solution for enhancing ASR performance. Experiments demonstrate that pre-training on 70,000-hour unlabeled data and fine-tuning on merely 50-hour labeled data yield a lightweight but powerful ASR model. It outperforms Whisper Large-v3 and commercial ASR systems on real-world data. Our code and models will be open-sourced to facilitate research in low-resource ASR.
title VietASR: Achieving Industry-level Vietnamese ASR with 50-hour labeled data and Large-Scale Speech Pretraining
topic Audio and Speech Processing
Artificial Intelligence
Computation and Language
Sound
url https://arxiv.org/abs/2505.21527