The Fante Multispeaker Audio Transcribed dataset is a collection of speech recordings and their transcriptions in Fante, a widely spoken dialect of the Akan language in Ghana. The dataset is designed for training and evaluating automatic speech recognition (ASR) models and other natural language processing (NLP) applications.

Dataset Details

Source: The dataset is derived from the Financial Inclusion Speech Dataset, which focuses on financial conversations in Fante.
Format: The dataset consists of audio recordings (.wav) and corresponding transcriptions (.txt or .csv).
Speakers: Multiple speakers contribute to the dataset, making it useful for speaker-independent ASR models.
Domain: Primarily focused on financial and general conversations.

Splits

The dataset is divided as follows:

Train Set: 90% of the data
Test Set: 10% of the data

Use Cases

This dataset is useful for:

Training and evaluating ASR models for Fante.
Developing Fante language models and NLP applications.
Linguistic analysis of Fante speech.

Usage

To use this dataset in your Hugging Face project, you can load it as follows:

from datasets import load_dataset

dataset = load_dataset("michsethowusu/fante_multispeaker_audio_transcribed")

License

Refer to the original dataset repository for licensing details: Financial Inclusion Speech Dataset.

Acknowledgments

This dataset is based on the work by Ashesi-Org. Special thanks to contributors who helped in data collection and annotation. I am only making it more accessible for machine learning.

Citation

If you use this dataset in your research or project, please cite it appropriately:

@misc{financialinclusion2022,
  author = {Asamoah Owusu, D., Korsah, A., Quartey, B., Nwolley Jnr., S., Sampah, D., Adjepon-Yamoah, D., Omane Boateng, L.},
  title = {Financial Inclusion Speech Dataset},
  year = {2022},
  publisher = {Ashesi University and Nokwary Technologies},
  url = {https://github.com/Ashesi-Org/Financial-Inclusion-Speech-Dataset}
}

Downloads last month: 29