| --- |
| title: README |
| emoji: 🐠 |
| colorFrom: green |
| colorTo: yellow |
| sdk: static |
| pinned: false |
| --- |
| |
| # NaathNLP |
|
|
| Open NLP infrastructure for **Thok Naath** (the Nuer language) — built by volunteers working toward language preservation and revitalization. |
|
|
| Nuer is spoken by millions of people across South Sudan, Ethiopia, and diaspora communities, but remains a low-resource language with almost no native NLP tooling. NaathNLP exists to change that — building translation models, parallel corpora, and language tools with quality reviewed by native speakers at every step. |
|
|
| ## What we're building |
|
|
| - **Translation** — a fine-tuned NLLB model for English–Nuer translation |
| - **Corpus** — a large-scale English–Nuer parallel corpus (1M+ pairs) |
| - **Chatbot** — a pivot-pipeline conversational prototype |
| - **ASR** — automatic speech recognition for spoken Nuer |
| - **TTS** — text-to-speech for Nuer, to support learners and low-literacy speakers |
| - **Reasoning** — working toward a native Nuer reasoning model that thinks in Nuer directly, rather than pivoting through English |
|
|
| ## Why this matters |
|
|
| Most language models have never seen meaningful amounts of Nuer text or speech. Without deliberate investment, low-resource languages like Naath risk falling further behind as AI tools become part of everyday life — for education, communication, and access to information. NaathNLP is a volunteer effort to make sure Nuer speakers aren't left out of that shift, and to help preserve the language for future generations. |
|
|
| ## Get involved |
|
|
| This is volunteer-driven work, and contributions are welcome — whether that's data, native-speaker review, compute, or code. Reach out if you'd like to help. |
|
|
| ngunartaban@gmail.com |
| github.com/bielng |
