|
--- |
|
annotations_creators: [] |
|
language: |
|
- ha |
|
language_creators: |
|
- expert-generated |
|
license: |
|
- cc-by-sa-4.0 |
|
multilinguality: |
|
- monolingual |
|
pretty_name: BibleTTS Hausa |
|
size_categories: |
|
- 10K<n<100K |
|
source_datasets: |
|
- original |
|
tags: |
|
- bible |
|
task_categories: |
|
- automatic-speech-recognition |
|
- text-to-speech |
|
task_ids: [] |
|
--- |
|
|
|
# Dataset Card for BibleTTS Hausa |
|
|
|
## Table of Contents |
|
- [Table of Contents](#table-of-contents) |
|
- [Dataset Description](#dataset-description) |
|
- [Dataset Summary](#dataset-summary) |
|
- [Languages](#languages) |
|
- [Dataset Structure](#dataset-structure) |
|
- [Data Fields](#data-fields) |
|
- [Data Splits](#data-splits) |
|
- [Additional Information](#additional-information) |
|
- [Dataset Curators](#dataset-curators) |
|
- [Licensing Information](#licensing-information) |
|
- [Citation Information](#citation-information) |
|
|
|
## Dataset Description |
|
|
|
- **Homepage:** https://masakhane-io.github.io/bibleTTS/ |
|
- **Repository:** http://www.openslr.org/129/ |
|
- **Paper:** https://arxiv.org/abs/2207.03546 |
|
|
|
### Dataset Summary |
|
|
|
BibleTTS is a large high-quality open Text-to-Speech dataset with up to 80 hours of single speaker, studio quality 48kHz recordings. |
|
This is a Hausa part of the dataset. Aligned hours: 86.6, aligned verses: 40,603. |
|
|
|
### Languages |
|
|
|
Hausa |
|
|
|
## Dataset Structure |
|
|
|
### Data Fields |
|
|
|
- `audio`: audio path |
|
- `sentence`: transcription of the audio |
|
- `locale`: always set to `ha` |
|
- `book`: 3-char book encoding |
|
- `verse`: verse id |
|
|
|
### Data Splits |
|
|
|
- `dev`: Book of Ezra (264 verses) |
|
- `test`: Book of Colossians (124 verses) |
|
- `train`: all other books (40215 verses) |
|
|
|
## Additional Information |
|
|
|
*See [this notebook](https://github.com/seads-org/hausa-speech-recognition/blob/6993c5c74379c93a2416acac6126b60ce6e52df8/notebooks/prepare_bible_dataset.ipynb) for the code on how the dataset was processed. |
|
|
|
### Dataset Curators |
|
|
|
The dataset uploaded by [vpetukhov](https://github.com/VPetukhov/) who is not connected to the dataset authors. Please, see the project page for more info. |
|
|
|
### Licensing Information |
|
|
|
The data is released under a commercial-friendly [CC-BY-SA](https://creativecommons.org/licenses/by-sa/4.0/) license. |
|
|
|
### Citation Information |
|
|
|
Meyer, Josh, et al. "BibleTTS: a large, high-fidelity, multilingual, and uniquely African speech corpus." arXiv preprint arXiv:2207.03546 (2022). |
|
|
