Logo Lanfrica
  • Accueil
  • Atlas
  • Analyses
  • Documentation
  • Sign in

© 2026 Lanfrica. Tous droits réservés. Tous les droits d'auteur des ressources affichées sur le site Web Lanfrica appartiennent aux détenteurs de droits d'auteur d'origine, sauf indication contraire explicite.

FuaadBashi/speech-to-text-AI-models-for-the-Somali-language

Domaine:

natural language processing

Type de record:

modelproject
Créateur:
Fua
Hôte:
Train or adapt a speech-to-text model capable of accurately transcribing Somali audio. # Somali Speech Recognition & Cloud Infrastructure **Full-Stack AI/DevOps Assessment Project: Production-Ready Somali ASR with Automated Cloud Deployment** > **Assessment Status**: ✅ All requirements met for both Part A (AI/ASR) and Part B (DevOps/Infrastructure) --- ## 📋 Table of Contents - Overview - Assessment Results - Part A: Somali ASR Training - Part B: Cloud Infrastructure Automation - Quick Start - Repository Structure - Technical Stack - Documentation - Troubleshooting - License --- ## 🎯 Overview This repository demonstrates end-to-end AI/DevOps capabilities through two integrated components: ### Part A: AI/Machine Learning Train and evaluate state-of-the-art speech recognition models for the Somali language, achieving production-grade accuracy with comprehensive WER evaluation. ### Part B: DevOps/Infrastructure Automate deployment of production-ready cloud infrastructure using Terraform, supporting the full ML lifecycle from training to inference. **Key Achievement**: Exceeded assessment targets in both AI performance (7.41% WER vs 15-20% target) and infrastructure completeness (28 cloud resources automated). --- ## 🏆 Assessment Results ### Part A: AI/Speech-to-Text ✅ | Requirement | Target | Achieved | Status | |------------|--------|----------|--------| | **Model Training** | Somali ASR model | Wav2Vec2 XLS-R 300M | ✅ **COMPLETE** | | **Dataset Preparation** | Unrestricted | 5-min verification clip + manifest | ✅ **COMPLETE** | | **WER Target** | 15-20% | **7.41%** (Overall: 9.09%) | ✅ **EXCEEDED** | | **Evaluation** | WER calculation | 129 segments, detailed metrics | ✅ **COMPLETE** | | **Documentation** | Methodology | CSV/JSON/TXT outputs | ✅ **COMPLETE** | **Result**: 🎉 **Target WER exceeded by 2x** (achieved 7.41%, target was 15-20%) ### Part B: DevOps/Infrastructure ✅ | Requirement | Target | Achieved | Status | |------------|--------|----------|--------| | **VPC & Networking** | VPC, Subnets, Security Groups | VPC + 2 subne …

Visit

github.com

Tasks

automatic speech recognitionspeech processing

Languages

Somali