Garnet-Owl/taura-datasets

Taura Datasets

This repository holds the source data for Taura 2.0, an open-source Kikuyu to English machine translation system.

The Kikuyu are Kenya's largest ethnic group, with 9 to 10 million speakers making up roughly 17 to 20% of the country's population. Despite this, Kikuyu remains severely underrepresented in modern NLP and translation tools. This dataset is part of an effort to change that.

Contents

Folder	Description
bible/	Kikuyu and English Bible PDFs used to extract parallel verse pairs

More data will be added here as the project grows, including seed dictionaries, monolingual corpora, and evaluation sets.

Languages

• Kikuyu (ISO 639-3: kik) — a Bantu language spoken primarily in Kenya's Central Province
• English (en)

Source

The Bible PDFs were sourced for research and low-resource NLP purposes. The Kikuyu Bible is a well-known parallel text that has been used in prior linguistic work on the language.

Parallel verse-level CSV files extracted from these PDFs are committed directly to the Taura 2.0 code repository under data/parallel/.

Usage

Download the PDFs and place them in the data/ directory of the Taura 2.0 project before running the preprocessing pipeline:

uv run python -m app.preprocessing.bible.orchestrator