aline-gassenn/MedDialog-Audio_v2

MedDialogue-Audio (Subset: Noise-Free - Parquet)

Description:
This dataset consists of clean audio files in Parquet format, intended for research in Automatic Speech Recognition (ASR).
It includes only the original, noise-free audio samples synthesized from medical dialogues.

The full version of the dataset, which includes white noise and hospital background noise variations in .wav format, is available at:
https://huggingface.co/datasets/aline-gassenn/MedDialog-Audio

---

Metadata

license: cc-by-nc-4.0
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-*
dataset_info:
  features:
    - name: audio
      dtype: audio
  splits:
    - name: train
      num_bytes: 21,068 
      num_examples: 21068
  download_size: ~22.3 GB
  dataset_size: ~23.5 GB

---

License

License: CC BY-NC 4.0

---

Citation

If you use this dataset, please cite:

@misc{med_dialogue_audio_v2_2025,
    author       = {Gassenn, Aline Elí and Teodoro, Douglas and Andrade, Luis G M and Rodrigues-Jr, Jose F.},
    title        = { MedDialog-Audio_v2 },
    year         = 2025,
    url          = { https://huggingface.co/datasets/aline-gassenn/MedDialog-Audio_v2 },
    doi          = { 10.57967/hf/5958 },
    publisher    = { Hugging Face }
}