README.md

metadata

language:
  - pl
license: cc0-1.0
size_categories:
  - 10K<n<100K
task_categories:
  - text-to-speech
  - automatic-speech-recognition
pretty_name: The MC Speech Dataset
dataset_info:
  features:
    - name: audio
      dtype:
        audio:
          sampling_rate: 44100
    - name: transcript
      dtype: string
    - name: id
      dtype: string
  splits:
    - name: train
      num_bytes: 6985316587.668
      num_examples: 24018
  download_size: 6174661195
  dataset_size: 6985316587.668
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-*

This is public domain speech dataset consisting of 24018 short audio clips of a single speaker reading sentences in Polish. A transcription is provided for each clip. Clips have total length of more than 22 hours.

Texts are in public domain. The audio was recorded in 2021-22 as a part of my master's thesis and is in public domain.

If you use this dataset, please cite:

@masterthesis{mcspeech,
  title={Analiza porównawcza korpusów nagrań mowy dla celów syntezy mowy w języku polskim},
  author={Czyżnikiewicz, Mateusz},
  year={2022},
  month={December},
  school={Warsaw University of Technology},
  type={Master's thesis},
  doi={10.13140/RG.2.2.26293.24800},
  note={Available at \url{http://dx.doi.org/10.13140/RG.2.2.26293.24800}},
}

More info about the dataset can be found at https://github.com/czyzi0/the-mc-speech-dataset

Also, if you find this resource helpful, kindly consider leaving a like.