README.md

metadata

license: cc0-1.0
task_categories:
  - automatic-speech-recognition
language:
  - en
  - de
  - fr
  - it
  - ru
  - es
tags:
  - speech
  - phonetics
  - phoneme-recognition
  - acoustic-modeling
pretty_name: Common Phone
size_categories:
  - 10K<n<100K

Dataset Card for Common Phone

This corpus aims to provide a basis for Machine Learning (ML) researchers and enthusiasts to train and test their models against a wide variety of speakers, hardware/software ecosystems and acoustic conditions to improve generalization and availability of ML in real-world speech applications. The current version of Common Phone comprises 116,5 hours of speech samples, collected from 11.246 speakers in 6 languages.

Common Phone has been used as the primary corpus for the PhD thesis of Philipp Klumpp: Phonetic Transfer Learning from Healthy References for the Analysis of Pathological Speech. The respective Wav2Vec2 model can also be found on Hugging Face.

It also served as a foundation for highly robust production-grade speech and phoneme recognition models.

Dataset Details

For details, check out our peer-reviewed publication Common Phone: A Multilingual Dataset for Robust Acoustic Modelling, which was presented at LREC 2022.

If you use Common Phone for your research, please cite our work:

 @inproceedings{
  klumpp2022common,
  title={Common Phone: A Multilingual Dataset for Robust Acoustic Modelling},
  author={Klumpp, Philipp and Arias, Tomas and P{\'e}rez-Toro, Paula Andrea and Noeth, Elmar and Orozco-Arroyave, Juan},
  booktitle={Proceedings of the Thirteenth Language Resources and Evaluation Conference},
  pages={763--768},
  year={2022}
}

• Curated by: Philipp Klumpp
• Language(s) (NLP): English, German, Italian, Spanish, French, Russian
• License: Creative Commons Zero v1.0 Universal

Dataset Sources

The dataset sources, including mp3/wav files and TextGrids, are also available via Zenodo.

• Repository: Common Phone on Zenodo

Uses

Why Common Phone?

• Large number of speakers and varying acoustic conditions to improve robustness of ML models
• Time-aligned IPA phonetic transcription for every audio sample
• Gender-balanced and age-group-matched (equal number of female/male speakers in every age group)
• Support for six different languages to leverage multi-lingual approaches
• Original MP3 files plus standard WAVE files

Source Data

Common Phone was derived from Common Voice, revision 7.0. More information can be found on the project's website.

Annotations

Common Phone provides automatically generated phonetic annotation with alignment for each spoken text sample.