Skip to content

Dataset · formospeech

Hakkaradio News Audio Corpus

formospeech/hakkaradio_news

Audio dataset of news recordings with text transcripts for rare dialects.

What's inside

The corpus includes audio and text ranging from 1K to 10K samples (total duration around 350 hours). Data is split into subsets (dapu, hailu, raoping, sixian, zhaoan) specifying sample counts and hours for each. Dataset access is public (Gated).

What you can do with it

  • Training and adapting ASR systems for dialects
  • Speech technologies and audio processing
Size
1K–10K rows
Downloads this month
22.9K
Likes
0
Access
on request on Hugging Face

How to load it

from datasets import load_dataset

ds = load_dataset("formospeech/hakkaradio_news")