---
dataset_info:
  config_name: Hakka_Sixian
  features:
  - name: id
    dtype: string
  - name: duration
    dtype: float64
  - name: hanzi
    dtype: string
  - name: hanzi_cln
    dtype: string
  - name: pinyin
    dtype: string
  - name: pinyin_cln
    dtype: string
  - name: ipa
    dtype: string
  - name: ipa_cln
    dtype: string
  - name: mandarin
    dtype: string
  - name: lang_group
    dtype: string
  - name: lang_group_en
    dtype: string
  - name: speaker
    dtype: string
  - name: mismatched_trs
    dtype: bool
  - name: audio
    dtype: audio
  splits:
  - name: train
    num_bytes: 42602180926
    num_examples: 88262
  download_size: 40131987666
  dataset_size: 42602180926
configs:
- config_name: Hakka_Sixian
  data_files:
  - split: train
    path: Hakka_Sixian/train-*
task_categories:
- text-to-speech
language:
- zh
- hak
size_categories:
- 10K<n<100K
---
# hat_asr_sixian_broadcast_clean_r

This dataset is a denoised `-R` variant of [formospeech/hat_asr_sixian_broadcast_clean](https://huggingface.co/datasets/formospeech/hat_asr_sixian_broadcast_clean).

## Summary

- Subset: `Hakka_Sixian`
- Dialect: `客語四縣`
- Train samples: `88262`
- Audio: denoised `24 kHz` WAV

## TRAIN

| Subset | lang_group | hours | n_utts | n_chars | secs/utt | chars/sec |
|---|---|---|---|---|---|---|
| Hakka_Sixian | 客語_四縣 | 231.77 | 88,262 | 3,629,753 | 9.45 | 4.35 |
| **Total** | **-** | **231.77** | **88,262** | **3,629,753** | **9.45** | **4.35** |

## Processing

1. Start from the original `formospeech/hat_asr_sixian_broadcast_clean` train split.
2. Run speech enhancement with [sarulab-speech/sidon-v0.1](https://huggingface.co/sarulab-speech/sidon-v0.1/tree/main).
3. Replace only the `audio` field with denoised audio.
4. Keep the remaining metadata fields aligned with the source dataset.