Logo Lanfrica

Joshmaina/ubuntu-ntu

Domain:

natural language processing

Record type:

softwareproject
Creator:
Jos
Host:
Offline-first, tone-aware mobile app for learning and preserving African languages. Community-owned language archive. # UBUNTU-NTU **Open infrastructure for learning — and preserving — African languages.** *Offline-first · Tone-aware · Community-owned* -lightgrey) --- ## What this is Roughly one third of the world's living languages are African. Digital language technology serves almost none of them. UBUNTU-NTU is a mobile application that teaches African languages the way they are actually spoken — **by ear, by tone, and in cultural context** — and builds a permanent open archive of those languages in the process. The two goals are the same activity. Every recording that teaches a learner also preserves a language. > ⚠️ **Current status: pre-development.** The documentation set is complete. No code > has been written yet. See the roadmap. --- ## Why it is different **It works with no internet.** Not "works offline too" — offline is the *primary* mode. Lessons download once as a few megabytes. Checking an answer, playing audio, recording your voice, and analysing your pitch all happen on the device. A learning session transfers zero bytes. **It runs on a phone and in a browser.** A mobile app (Android-first) for learners on the continent, where guaranteed offline and data economy matter most — and an installable web app for the diaspora, desktop learners, and anyone trying it without installing anything. Both work offline; both share the same scheduling engine, sync protocol, and pitch analysis. → Platform strategy **It teaches tone, visibly.** Many African languages are tonal: the same syllables at a different pitch mean a different word. Being told "your tone was wrong" is useless. So we draw your pitch against the native speaker's, normalised so a deep voice and a high voice are compared fairly. You can *see* the difference, and see it close. **It never drops you into a cold sentence.** Our core exercise, *Scaffolded Sentence Deconstruction*, removes cognitive load one layer at a time: ```text 1. ANCHOR "I am going to the market." ← meaning first …