gmahia/swahili-civic-nlp

Swahili Civic NLP

Swahili language NLP dataset for civic accountability tasks including constitutional rights, government services, and public interest questions. Supports NLP models for East Africa's 200M+ Swahili speakers.

Dataset Description

This dataset is part of the East Africa AI coordination infrastructure — a portfolio of open-source tools addressing structural gaps in AI coverage for Kenya and East Africa.

Intended Use

• Training and fine-tuning language models for East African contexts
• Powering AI agents in the MCP server ecosystem
• Research into AI applications for the Global South
• Building civic technology tools for Kenya's 47 counties

Data Sources

Data is sourced from public government records, open civic databases, and structured research. All data is labeled with source and confidence indicators.

Citation

If you use this dataset, please cite:

@dataset{mahia2024swahilicivicnlp,
  title={Swahili Civic NLP},
  author={Gabriel Mahia},
  year={2024},
  publisher={HuggingFace},
  url={https://huggingface.co/datasets/gmahia/swahili-civic-nlp}
}

Related Resources

• MCP Servers for East Africa
• GitHub Portfolio
• PyPI Packages

Use this dataset with the coordination stack

• Evaluate any model on Swahili agent tasks: kipimo — pip install kipimo · leaderboard
• Run the MCP servers this data serves: pip install reli-cli && reli list — 33 servers on the official MCP Registry under io.github.gabrielmahia
• Route tasks across servers: africa-coord-bus — offline-first event bus with a built-in Kenya routing table
• All East Africa datasets: huggingface.co/gmahia

Open weights welcome: these resources are model-agnostic by design — closed APIs, open-weight models (Kimi, GLM, DeepSeek, Nemotron, Llama), and small distilled models score on equal terms.