Datasets:

formospeech
/

hat_asr_sixian_broadcast

dataset_info:
  config_name: hak_sx
  features:
    - name: id
      dtype: string
    - name: stem
      dtype: string
    - name: audio
      dtype: audio
    - name: duration
      dtype: float64
    - name: hanzi
      dtype: string
    - name: pinyin
      dtype: string
    - name: ipa
      dtype: string
    - name: mandarin
      dtype: string
    - name: lang_code
      dtype: string
    - name: lang_name
      dtype: string
    - name: channel
      dtype: 'null'
    - name: speaker
      dtype: 'null'
  splits:
    - name: train
      num_bytes: 27680861034
      num_examples: 79253
  download_size: 27427769901
  dataset_size: 27680861034
configs:
  - config_name: hak_sx
    data_files:
      - split: train
        path: hak_sx/train-*

TRAIN

Subset	lang_name	hours	n_utts	n_chars_in_utts	secs/utt	chars/sec	n_sents	n_chars_in_sents
hak_sx	Hakka_Sixian	237.66	79,253	3,637,796	10.80	4.25	0	0
Total	-	237.66	79,253	3,637,796	10.80	4.25	0	0