Researching and Experimenting on NLP in African contexts
# Kalenjin–Kiswahili Speech & Translation AI
This project demonstrates an end-to-end **Natural Language Processing (NLP)** and **Text-to-Speech (TTS)** pipeline using a **Kalenjin–Kiswahili parallel corpus**. It showcases how to load linguistic datasets, generate speech, perform simple translations, and create basic AI-generated videos.
---
## Project Overview
This notebook (`pochon.ipynb`) walks through a multi-stage AI workflow:
- Loading and exploring a bilingual dataset
- Generating speech using AI models
- Translating between Kiswahili and Kalenjin
- Creating simple videos from audio + images
The project demonstrates **AI applications for African languages**, including:
- Language preservation
- Voice assistants
- Education tools
- Accessibility technologies
---
## Tech Stack
- **Python**
- **Transformers (Hugging Face)** – NLP & TTS models
- **PyTorch** – GPU acceleration
- **Pandas** – Data handling
- **SoundFile** – Audio processing
- **MoviePy** – Video generation
---
## Installation
```bash
pip install transformers datasets soundfile accelerate -q
pip install pandas pyarrow rustbpe tiktoken -q
pip install moviepy -q