metadata
dataset_info:
features:
- name: audio
dtype: audio
- name: transcript
dtype: string
- name: za_asr
dtype: string
- name: WER
dtype: int64
- name: CER
dtype: int64
- name: lower_WER
dtype: int64
- name: lower_CER
dtype: int64
- name: source
dtype: string
- name: audio_id
dtype: string
splits:
- name: train
num_bytes: 110015985.75
num_examples: 1002
download_size: 109486059
dataset_size: 110015985.75
configs:
- config_name: default
data_files:
- split: train
path: data/train-*
- Not supported long audio (tested with 20min)