Logo Lanfrica

Garnet-Owl/taura

Domaine:

natural language processing

Type de record:

model
Créateur:
Gar
Hôte:
A project to build machine translation model for translating Kikuyu to English # Taura - Kikuyu-English Translation Project The name "taura" is derived from the Kikuyu word 'Taũra', which means 'Translate', which is exactly what this project aims to do. ## Project Overview Taura is a machine translation model for translating text between Kikuyu and English languages. Kikuyu is a Bantu language spoken by the Kikuyu people of Kenya. It's the most widely spoken language in Kenya, after English and Swahili. ## Project Structure ``` taura/ ├── data/ # Data directory │ ├── raw/ # Raw collected data │ ├── processed/ # Processed and tokenized data │ ├── English20Kikuyu20Pairs2029.xlsx # Sample dataset │ └── external/ # External datasets ├── models/ # Model checkpoints and configurations ├── notebooks/ # Jupyter notebooks for experimentation ├── src/ │ ├── data/ # Scripts for data collection, preparation │ │ ├── explore_data.py # Script to explore the dataset │ │ ├── preprocess.py # Data preprocessing script │ │ └── tokenize.py # Tokenization script │ ├── features/ # Feature engineering │ ├── models/ # Model definition and training scripts │ │ ├── translation_model.py # Model architecture definition │ │ ├── train.py # Training script │ │ ├── evaluate.py # Evaluation script │ │ └── translate.py # Inference script for translation │ └── visualization/ # Scripts for visualizing results ├── results/ # Evaluation results and outputs ├── tests/ # Unit tests ├── .gitignore # Git ignore file ├── LICENSE # License file ├── README.md # Project documentation (this file) └── requirements.txt # Project dependencies ``` ## Getting Started ### Prerequisites - Python 3.12+ - PyTorch 1.9+ - Transformers 4.20+ ### Installation 1. Clone the repository: …