g-group-ai-lab/vi-asr-pubadmin-test

Vietnamese ASR Test Set - Public Administration

A Vietnamese speech-recognition benchmark for the public administration domain (Hành chính công), released by G-Group AI Lab.

Audio is real-world Vietnamese speech covering administrative procedures, paperwork and licensing guidance, civil records — dense in place names and legal terminology.

Listen & explore

Every utterance is playable inline in the viewer above — hit play on any row to stream the clip. For full-text search across transcripts, duration filters and a DuckDB SQL console over the whole split, open Data Studio →.

At a glance

Utterances	515
Total duration	0.47 hours
Audio	MP3, 16 kHz, mono
Language	Vietnamese (vi)
Split	test
Task	Automatic speech recognition

Usage

from datasets import load_dataset

ds = load_dataset("g-group-ai-lab/vi-asr-pubadmin-test", split="test")
print(ds[0]["text"])
print(ds[0]["audio"]["sampling_rate"])  # 16000

Stream it without downloading the whole set:

ds = load_dataset("g-group-ai-lab/vi-asr-pubadmin-test", split="test", streaming=True)
print(next(iter(ds))["text"])

Fields

Column	Type	Description
id	string	Utterance id (matches the source audio file stem)
audio	Audio(16 kHz)	Mono MP3, embedded as-is (no re-encoding)
text	string	Reference transcript
duration	float32	Audio duration in seconds

Related test sets

Domain	Dataset
Technology	g-group-ai-lab/vi-asr-tech-test
Education	g-group-ai-lab/vi-asr-edu-test
Public administration	g-group-ai-lab/vi-asr-pubadmin-test
Finance	g-group-ai-lab/vi-asr-finance-test

License

Released under CC BY-NC 4.0 — free for non-commercial use with attribution.

Citation

@misc{vi_asr_pubadmin_test_2026,
  title  = {Vietnamese ASR Test Set - Public Administration},
  author = {G-Group AI Lab},
  year   = {2026},
  url    = {https://huggingface.co/datasets/g-group-ai-lab/vi-asr-pubadmin-test}
}