metadata

dataset_info:
  features:
    - name: audio
      dtype:
        audio:
          sampling_rate: 44100
    - name: text
      dtype: string
  splits:
    - name: train
      num_bytes: 5938096417.625
      num_examples: 19947
  download_size: 4622425790
  dataset_size: 5938096417.625
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-*
license: cc-by-4.0
task_categories:
  - text-to-speech
language:
  - wo
pretty_name: anta_women_tts
size_categories:
  - 10K<n<100K

Anta Women TTS

Dataset Description

This is a cleaned version of the Wolof TTS dataset by GalsenAI. We extracted the female voice, denoised it and enhanced it with the Resemble Enhance library. We also cleaned up the annotations by removing special characters, emojis, Arabic and Russian characters. We've corrected a few annotation errors, but there are potentially many more to come. Some lines and audios judged not qualitative enough have been removed from the dataset, reducing its size to 19947 annotated audios.

Speaking time:

The overall speaking time is: 18h41mn46s

speaking time before cleaning: 18h 47mn 19s.

The text dataset comes from news websites, Wikipedia and self curated text.

Citation

You can access the project paper at https://arxiv.org/abs/2104.02516.
If you work on the dataset, please cite the authors below.

@dataset{thierno_ibrahima_diop_2021_4498861,
  author       = {Thierno Ibrahima Diop and
                  Demba AW and
                  Ami jaane and
                  Mamadou Badiane},
  title        = {WOLOF TTS(Text To Speech) Data},
  month        = feb,
  year         = 2021,
  publisher    = {Zenodo},
  version      = 1,
  doi          = {10.5281/zenodo.4498861},
  url          = {https://doi.org/10.5281/zenodo.4498861}
}

Acknowledgment

This work was carried out by Derguene and was made possible thanks to computing infrastructure support from Caytu Robotics. Many thanks to them and especially to Abdoulaye Faye for this support.