maanka2/fleurs-somali
FLEURS Somali FLEURS Somali is a processed Somali speech dataset derived from the Somali portion of the Google FLEURS corpus. The dataset is designed for Automatic Speech Recognition (ASR), Speech-to-Text (STT), and speech technology research. Audio samples have been enhanced through noise reduction and normalization while preserving the original speech content and transcriptions. Features Feature Type audio Audio transcription String… See the full description on the dataset page: https://huggingface.co/datasets/maanka2/fleurs-somali.
FLEURS Somali
FLEURS Somali is a processed Somali speech dataset derived from the Somali portion of the Google FLEURS corpus. The dataset is designed for Automatic Speech Recognition (ASR), Speech-to-Text (STT), and speech technology research.
Audio samples have been enhanced through noise reduction and normalization while preserving the original speech content and transcriptions.
Dataset Description
- Creator: maanka2
- Source Dataset: Google FLEURS (Somali)
- Language: Somali (so)
- Sampling Rate: 16 kHz
- Speaker Type: Multiple speakers
- Task: Automatic Speech Recognition (ASR)
Features
Processing Pipeline
The dataset was processed using:
- Noise Reduction
- Audio Normalization
- 16 kHz Audio Standardization
These steps improve audio consistency and reduce background noise for ASR training.
Intended Uses
- Automatic Speech Recognition
- Speech-to-Text Systems
- Acoustic Model Training
- Speech Foundation Models
- Somali Language Technology
- Low-Resource Language Research
Dataset Structure
The dataset contains Somali speech recordings paired with their corresponding transcriptions.
Limitations
- Derived from the Somali subset of Google FLEURS.
- Recording conditions may vary across speakers.
- Not intended for speaker identification tasks.
Citation
If you use this dataset in research or production systems, please cite both this dataset and the original Google FLEURS dataset.
Acknowledgements
This dataset is based on the Google FLEURS Somali corpus and has been further processed to support Somali Automatic Speech Recognition research and development.
