Logo Lanfrica

AnanduMadhu01/koze_kreol

Domaine:

natural language processing

Type de record:

software
Créateur:
Ana
Hôte:
MUA kreol # Morisien CAPT API — Full Technical Documentation A REST API for **Computer-Assisted Pronunciation Training (CAPT)** for the Morisien (Creole) language. It accepts audio recordings and returns pronunciation scores and phoneme-level feedback. Base URL: `localhost` --- ## Table of Contents 1. System Architecture Overview 2. The ASR Model — What It Is and How It Works 3. Audio Processing Pipeline 4. Phonemizer Pipeline 5. GOP-SF Scoring Algorithm 6. Identity Check — Did They Say the Right Word? 7. End-to-End Workflow (Step by Step) 8. API Endpoints 9. Audio Requirements 10. Running the Server 11. Dependencies --- ## 1. System Architecture Overview ``` ┌─────────────────────────────────────────────────────────────────┐ │ FastAPI Server (app.py) │ │ │ │ POST /user ──────────────────────────────────────────────┐ │ │ POST /admin ─────────────────────────────────────────┐ │ │ │ GET /lexicon │ │ │ │ GET /health │ │ │ └────────────────────────────────────────────────────────┼───┼───┘ │ │ ┌────────────────────────────────────┘ │ ▼ ▼ ┌──────────────────┐ ┌────────────────────┐ │ AudioProcessor │ │ CreoleCAPTEngine │ │ (preprocessing) │ │ (assess method) │ └────────┬─────────┘ └────────┬───────────┘ │ │ ▼ ▼ ┌──────────────────┐ ┌─────────────────────────┐ │ MultilingualPhone│ │ CreolePhonemizerPipeline │ │ meASR (Wav2Vec2) │◄─────────────│ (word → IPA phonemes) │ └──────────────────┘ └─────────────────────────┘ │ ▼ ┌──────────────────┐ │ GOP-SF Scoring │ │ (forward-backward│ │ CTC alignment) │ └──────────────────┘ …