license: cc-by-4.0
language:
- zu
- en
size_categories:
- 1K<n<10K
🧠 MetaWell Code-Switching Dataset
The MetaWell Code-Switching Dataset is the first large-scale African dataset built for training and evaluating code-switching between isiZulu and English.
It contains paired audio + text examples representing natural language use in South African daily life, media, and conversation. The dataset enables development of speech-to-text, translation, and language understanding models for African multilingual environments.
📊 Dataset Structure Language Samples Type Content isiZulu ~1 000 Text + Audio Native isiZulu speech and transcripts English ~1 000 Text + Audio Native English speech and transcripts Code-Switched (Zulu-English) ~400 Text + Audio Mixed-language examples
Each entry includes:
text/ transcript
audio/ recording (.wav)
metadata (language, speaker, emotion, tone, etc.)
⚙️ Intended Use
Code-switching recognition and translation
African-language speech recognition
Multilingual chatbots and assistants
AI education and research in Africa
📂 File Organization MetaWell_AI_Dataset/ ├── text/ │ ├── zulu/ │ ├── english/ │ └── codeswitch/ ├── audio/ │ ├── zulu/ │ ├── english/ │ └── codeswitch/ └── metadata.csv
🏷️ License
CC-BY 4.0 — You may use, share, or adapt this dataset with attribution to MetaWell AI.
🧩 Citation
If you use this dataset, please cite:
@dataset{moremoholo_metawell_2025, title = {MetaWell Code-Switching Dataset (Zulu–English)}, author = {Moremoholo, Thabo}, year = {2025}, url = {https://huggingface.co/metawell-ai/metawell_codeswitch_dataset} }
🌍 About MetaWell AI
MetaWell AI is an African AI initiative focused on bridging human and machine communication through multilingual, code-switching, and culturally aware models.
Homepage: https://meta-vault-1daccafc.base44.app
Contact: thabo1.moremoholo@gmail.com