Vietnamese Automatic Speech Recognition: A Revisit

2026-03-15

AI Summary

This work addresses the persistent scarcity of high-quality speech recognition data for low-resource languages such as Vietnamese, where existing open-source datasets often suffer from inconsistent annotations and substantial noise. To overcome these limitations, the authors propose a generalizable data aggregation and preprocessing pipeline that integrates multiple open-source sources. By applying rigorous audio-text alignment, noise filtering, dataset balancing, and preservation of word-level timestamps, they construct a high-quality, diverse, and balanced 500-hour Vietnamese ASR dataset. This resource significantly enhances the availability of reliable training and evaluation data for low-resource speech recognition, providing a robust foundation for developing and benchmarking state-of-the-art Vietnamese automatic speech recognition systems.

Technology Category

• Natural Language Processing: Speech
• Machine Learning: Large Multimodal Models (LMMs)
• Computer Vision: Language and Vision

Application Category

• Web Mining and Content Analysis: Web data integration and cleaning
• Economics, Online Markets and Human Computation: Data quality aspects of human-annotated datasets
• Search and Retrieval-Augmented AI: Multilingual and cross-lingual Web search

Abstract

Automatic Speech Recognition (ASR) performance is heavily dependent on the availability of large-scale, high-quality datasets. For low-resource languages, existing open-source ASR datasets often suffer from insufficient quality and inconsistent annotation, hindering the development of robust models. To address these challenges, we propose a novel and generalizable data aggregation and preprocessing pipeline designed to construct high-quality ASR datasets from diverse, potentially noisy, open-source sources. Our pipeline incorporates rigorous processing steps to ensure data diversity, balance, and the inclusion of crucial features like word-level timestamps. We demonstrate the effectiveness of our methodology by applying it to Vietnamese, resulting in a unified, high-quality 500-hour dataset that provides a foundation for training and evaluating state-of-the-art Vietnamese ASR systems. Our project page is available at https://github.com/qualcomm-ai-research/PhoASR.

Problem

• Automatic Speech Recognition
• low-resource languages
• dataset quality
• annotation inconsistency
• Vietnamese ASR

Innovation

• data aggregation
• preprocessing pipeline
• low-resource ASR
• word-level timestamps
• Vietnamese speech recognition