metadata
dataset_info:
features:
- name: id
dtype: string
- name: audio
dtype: audio
- name: text
dtype: string
splits:
- name: train
num_bytes: 1037543636.9388236
num_examples: 1690
download_size: 855276293
dataset_size: 1037543636.9388236
configs:
- config_name: default
data_files:
- split: train
path: data/train-*
license: cc0-1.0
Retrieved from https://huggingface.co/datasets/facebook/voxpopuli. Preprocessed for the experiments of Re-evaluating Minimum Bayes Risk Decoding for Automatic Speech Recognition.