---
annotations_creators:
  - machine-generated
language:
  - ngl
language_creators:
  - found
license:
  - mit
multilinguality:
  - monolingual
pretty_name: "{language_name} Speech-Text Dataset"
task_categories:
  - automatic-speech-recognition
  - text-to-speech
task_ids:
  - keyword-spotting
  - audio-intent-classification
---

# Lomwe Speech-Text Parallel Dataset

This dataset is a collection of aligned audio-text pairs in **Lomwe**, extracted from the [CMU Wilderness dataset](https://github.com/festvox/datasets-CMU_Wilderness). It is useful for tasks such as:

- Speech recognition (ASR)
- Text-to-speech (TTS)
- Language modeling for low-resource languages

## Dataset Structure

Each entry in the dataset contains:
- `audio`: A `.wav` file sampled at 16kHz
- `text`: A transcription of the spoken audio in Lomwe (digits removed)

### Example

| audio | text |
|-------|------|
| `B04___15_John________LONBSMN2DA_00001.wav` | `“Miyaano tti muteekho wa mpesa wawoona nave Attiitthaaka sika mwaneena a emata. Alliwa ennuuha erive enthambi ya miyaano ehinimma ezipaso nave ennaphaddera erive enthambi eyo enimma ezipaso wi yimme ezipaso ddawaata. Nyuwaano mookhalla aweella khalle wonanko enthambi yoophadderiwa ekaamba ya mawu kuulleelleemini. Khallani mwa miyaano nave miyaano kikhalle mwa nyuwaano.` |
| `B04___15_John________LONBSMN2DA_00002.wav` | `Enthambi khenimma ezipaso mwa weekhiiwa yaahikhalla vamuteekhoni wa mpesa.` |
| `B04___15_John________LONBSMN2DA_00003.wav` | `Emoddamoddave yeeyo, nyuwaano khamuwora wimma ezipaso mwaahikhalla mwa miyaano. “Miyaano tti muteekho wa mpesa nave nyuwaano sika enthambi.` |

## Language

- **Name**: Lomwe
- **Language Family**: Bantu (if applicable)
- **ISO 639-3**: `ngl` (change if needed)

## Source

This dataset is derived from:
- **CMU Wilderness**: A multilingual Bible corpus for speech research, published by Carnegie Mellon University.

## Preprocessing

- Numeric digits have been stripped from text.
- Audio and text are aligned by file ID.

## License

MIT License or original CMU Wilderness license (depending on your use case).

## Citation

@inproceedings{black2019cmu, title={CMU Wilderness Multilingual Speech Dataset}, author={Black, Alan W}, booktitle={ICASSP 2019-2019 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)}, year={2019}, organization={IEEE} }


## Maintainer

- [@michsethowusu](https://huggingface.co/michsethowusu)