metadata
license: apache-2.0
dataset_info:
features:
- name: audio
dtype:
audio:
sampling_rate: 16000
- name: transcription
dtype: string
splits:
- name: train
num_bytes: 29311788948.030003
num_examples: 182835
- name: test
num_bytes: 3181370673.15
num_examples: 20315
download_size: 29701298876
dataset_size: 32493159621.180004
Dataset instance structure
{'audio': {'path': '/path/to/wav.wav', 'array': array([wav numpy array]), dtype=float32), 'sampling_rate': 16000}, 'transcription': 'транскрипция'}
Dataset audio info
- 16000 Hz
- wav
- mono
- Russian speech from audiobooks
Citation
@misc{sova2021rudevices, author = {Zubarev, Egor and Moskalets, Timofey and SOVA.ai}, title = {SOVA RuDevices Dataset: free public STT/ASR dataset with manually annotated live speech}, publisher = {GitHub}, journal = {GitHub repository}, year = {2021}, howpublished = {\url{https://github.com/sovaai/sova-dataset}}, }