OpeyemiOsakuade/yoruba-g2p

Tone-aware Yoruba grapheme-to-phoneme tool (IPA + ASCII) for speech research and MFA. # 🇳🇬 Yoruba-G2P ### **Tone-Aware Yoruba Grapheme-to-Phoneme Toolkit (IPA + ASCII + MFA-Ready)** **Yoruba-G2P** is a fully deterministic Python package for converting **Yorùbá text → phoneme sequences**, with correct **tones**, **nasal handling**, **affricates**, and **labial-velars**. It outputs: - ✔ **IPA dictionary** - ✔ **ASCII-safe dictionary** (for MFA, ESPnet, Kaldi) - ✔ **Phoneset file** - ✔ **Lexicon statistics** - ✔ **CLI + Python API** - ✔ **Works on any Yoruba transcript** - ✔ No ML training required → fully rule-based + Epitran-backed --- ## Installation pip install epitran grapheme This version uses Unicode grapheme-aware processing to ensure correct handling of Yorùbá underdot+tone-marked vowels. ### From PyPI (recommended) ```bash pip install yoruba-g2p ```` ### From GitHub ```bash pip install git+github.com ``` --- ## 📣 Citation (for research) ## Citation If you use this work, please cite: ```bibtex @software{osakuade2026yoruba_g2p, author = {Osakuade, Opeyemi}, title = {Yoruba-G2P: A tone-aware grapheme-to-phoneme converter for Yorùbá}, year = {2026}, version = {v0.2.4}, doi = {10.5281/zenodo.19721964}, url = {github.com } ``` --- ## Quick Start (Python API) ```python from yoruba_g2p import YorubaG2P g2p = YorubaG2P() print(g2p.yoruba_word_to_ipa_phones("ọ̀yọ́")) print(g2p.ipa_phone_to_ascii("ọ̀mọ́")) ``` Output: ``` ['ɔ_L', 'j', 'ɔ_H'] ['O_L', 'm', 'O_H'] ``` Another example: ```python g2p.yoruba_word_to_ipa_phones("àwọn") ['a_L', 'w', 'ɔ_M', 'n'] ``` --- ## Command-line Interface (CLI) Convert a sentence: ```bash yoruba-g2p --ipa "àwọn ọmọ ń lọ" yoruba-g2p --ascii "àwọn ọmọ ń lọ" ``` Build lexicons from `.lab` transcripts: ```bash yoruba-g2p build-lexicon \ --lab-dir data/lab/train \ --out-dir dict/ ``` You will get: ``` dict/ipa.dict dict/ascii.dict dict/phoneset.txt dict/stats.json ``` All **MFA-ready**. --- ## What Yoruba-G2P Produces …