AndiswaMbonambi08/indabaX-health-nlp

Multilingual health Q&A retrieval system for African languages, built for the IndabaX x Zindi hackathon. TF-IDF baseline plus fine-tuned mT5-small model. Team BlackJack, 6th place on the public leaderboard. # IndabaX South Africa Multilingual Health Question Answering Hackathon ## 📌 Overview This repository contains the project developed during the **IndabaX South Africa Hackathon (July 2026)**, hosted on Zindi. Our team built a multilingual health question answering system designed for low-resource African languages. We ranked **6th out of 35 teams** in the competition. ## 🚀 Project Goals - Enable communities to access health information in multiple African languages. - Apply **Natural Language Processing (NLP)** and **Large Language Models (LLMs)** to low-resource contexts. - Demonstrate practical AI solutions for healthcare accessibility. ## 🛠 Tech Stack - **Python** - **NLP libraries**: Hugging Face Transformers, spaCy - **Data handling**: Pandas, NumPy - **Evaluation**: F1-score, accuracy metrics ## 📂 Repository Structure ├── data/ Datasets used for training/testing ├── notebooks/ Jupyter notebooks for experiments ├── src/ Source code for models and preprocessing ├── results/ Evaluation outputs, logs, and final submission └── README.md ## Project documentation ## ⚙️ How to Run 1. Clone the repository: ```bash git clone github.com ## Install dependencies pip install -r requirements.txt ## Run training python src/train.py ## Evaluate python src/evaluate.py 🏆 Results - Ranked 6th out of 35 teams. - Achieved strong performance on multilingual health QA tasks. - `results/final_submission.csv` — final predictions submitted to the public leaderboard (score 0.574223). 🙌 Acknowledgments - Zindi for hosting the hackathon. - IndabaX South Africa for organizing the event. - Team members and mentors who contributed to the project. ## Screenshots