metadata

task_categories:
  - automatic-speech-recognition
pretty_name: MangoSpeech
configs:
  - config_name: rozdympodcast
    data_files: data/rozdympodcast.parquet
  - config_name: opodcast
    data_files: data/opodcast.parquet
  - config_name: test
    data_files: data/test.parquet

The list of all subsets in the dataset

Each subset is generated splitting videos from given particular ukrainiam YouTube channel All subsets are in test split

"opodcast" subset is from channel "О! ПОДКАСТ"
"rozdympodcast" subset is from channel "Роздум | Подкаст"
"test" subset is just a small subset of samples

Loading a particular subset

>>> data_files = {"train": "data/<your_subset>.parquet"}
>>> data = load_dataset("Zarakun/youtube_ua_subtitles_test", data_files=data_files)
>>> data
DatasetDict({
    train: Dataset({
        features: ['audio', 'rate', 'duration', 'sentence'],
        num_rows: <some_number>
    })
})