Skip to content

Africa Bitext Builder

Africa Bitext Builder is a Python library for discovering, inspecting, and compiling sentence-aligned parallel and monolingual text corpora across 690+ African languages.

The library pairs African languages with English, French, Arabic, Portuguese, and Chinese, or directly aligns two African languages (e.g., Twi ↔ Yoruba, Hausa ↔ Amharic).

Core Principles

  1. Explicit Licensing Enforcement: Defaults strictly to verified Public Domain (PD) and Creative Commons (CC) translations.
  2. On-Demand Dataset Streaming: Data is downloaded only when requested from Hugging Face (AfriSpeech/africa-corpus) and cached locally.
  3. Canonical Alignment: Verses are matched deterministically using canonical scripture identifiers (GEN.1.1 to REV.22.21).
  4. Fast Discovery: Look up translation metadata, country distributions, and licensing across hundreds of dialects in constant time.

Installation

Install the latest published version directly from PyPI:

pip install africa-bitext-builder

uv add africa-bitext-builder

Next Steps

  • Language & License Discovery: Learn how to query ISO codes, verify Public Domain availability, and search by country or abbreviation.
  • Compiling Corpora: Configure parallel extraction, apply random sampling, and export CSV files.
  • API Reference: Detailed docstring references for LanguageRegistry and CorpusBuilder.