Logo Lanfrica

BMMUGOMBA/zimvoice

Domaine:

natural language processing

Type de record:

softwaredataset
Créateur:
BMM
Hôte:
he platform where Zimbabweans teach AI their language.** Zimbabwe Indigenous Language AI Speech Corpus Platform AI4I Challenge 2026 — Track 3: Development # ZimVoice 🌍 **The platform where Zimbabweans teach AI their language.** Zimbabwe Indigenous Language AI Speech Corpus Platform AI4I Challenge 2026 — Track 3: Development --- ## Quick Start (5 minutes) **Step 1: Install dependencies** ``` npm install ``` **Step 2: Create your .env file** Copy `.env.example` to `.env` and fill in your Firebase config values: ``` cp .env.example .env ``` Then edit `.env` and paste your Firebase config values from the Firebase Console. **Step 3: Run the development server** ``` npm run dev ``` Open localhost in Chrome. **Step 4: Run tests** ``` npm test ``` **Step 5: Deploy to Firebase** ``` npm run deploy ``` Your app will be live at zimvoice-zw.web.app --- ## Project Structure ``` src/ screens/ — The three main screens (Record, Verify, Dashboard) components/ — Shared UI components (TopBar, BottomNav) firebase/ — Firebase connection, auth, database, storage hooks/ — React hooks (useAudioRecorder) utils/ — Utilities (wordCounter, validators, offlineCache) ``` ## The Three Screens - **Record** — Language Guardians speak 3–10 words and type the transcript - **Verify** — Community votes on whether audio matches transcript - **Dashboard** — Live stats, National Language Map, AI Readiness Score --- ## Tech Stack & Dependencies - **Frontend:** React 18, Vite - **Backend:** Firebase (Auth, Firestore, Storage, Hosting) — the PWA calls the Firebase SDK directly; there is no separate custom API layer in the current build - **AI/ML:** Whisper-tiny (ONNX runtime, in-browser inference — no server-side model calls) - **Audio analysis:** meyda.js (SNR / audio quality scoring) - **Model training:** Google Colab (GPU, free tier) — fine-tuning cycle every 500 verified pairs - **Testing:** validators, PII detection, consensus logic Full dependency versions are listed in `package.json` — run `npm list` for the resolved tree. ## System Architecture See `architectu …