Ghayth-Bouzayeni/tunisian-voice-pipeline

Domain:

natural language processing

Record type:

software

Creator:

Gha

Host:

# 🗣️ Tunisian-Voice-Pipeline A **public speech-to-text API** designed for the **Tunisian Arabic dialect**, aggregating results from multiple ASR models and providing a correction dashboard for data collection and fine-tuning. This project is fully containerized and cloud-ready. --- ## 🎯 Project Objectives 1. Provide a **REST API** that accepts audio and returns aggregated transcriptions from multiple models: - Whisper Small - Whisper Medium - Wav2Vec2 (other models possible) 2. Collect user data (audio + corrected text) for **fine-tuning** Tunisian ASR models. 3. Deploy in **Kubernetes (Kind for local testing, AKS for production)**. 4. Use **OpenResty (Nginx + Lua)** as a reverse proxy for request routing and aggregation. 5. Provide a **dashboard** for admins to validate/correct transcriptions. 6. Use **Azure Blob Storage and Queue** for event-driven processing. --- ## 🧱 Architecture Overview *Swagger interface of the public API.* *Dashboard screenshot for admin transcription review.* ### Components | Component | Tech | Description | |-----------|------|-------------| | **Aggregation API** | Python (Flask) | Aggregates outputs from multiple ASR models and returns best transcription | | **Reverse Proxy** | OpenResty (Nginx + Lua) | Handles routing, ID assignment, and authentication | | **Queue & Blob** | Azure Storage | Manages messages for event-driven transcription collection and storage | | **Dashboard** | HTML/CSS/JS | Admin review and correction interface | | **Infrastructure** | Terraform + AKS / Kind | Cloud and local deployment manifests | | **Models (Docker)** | Whisper Small, Whisper Medium, Wav2Vec2 | Dockerized ASR models | --- ## 🌐 DockerHub Images - Whisper Small → ghaythbz/whisper-jdide - Whisper Medium → ghaythbz/whisper-medium-api - Aggregation API → ghaythbz/agg-api > Note: Models are **not stored in GitHub**. Pull images from DockerHub for local or cloud deployment. --- 🧠 Future Work Expand to more Tunisian dialect ASR models Fi …

rouatorjmen1/tunisian-hate-speech-pipeline

# Tunisian Dialect Hate Speech and Abuse Detection Using TunBERT This repository provides a comprehe

badIS-6/Tunisian-Agricultural-Data-Harmonization-GIS-Pipeline

A pipeline for harmonizing Tunisian agricultural datasets, resolving Arabic/French regional naming i

GHAYTH-LAB/choufli-hall-character-segmentation-tracking-yolov8

YOLOv8-based character segmentation for Choufli Hal, a popular Tunisian comedy TV series. This proje

Hybrid Pipeline for Building Arabic Tunisian Dialect-standard Arabic Neural Machine Translation Model from Scratch

Deep Learning is one of the most promising technologies compared to other methods in the context of

Methodological pipeline.

We designed a survey that combines questions about behavioral practices that could expose individ

swahili-voice/Swahili-voice-of-Msamiati

This project is a speech translation tool that translates a persons speech as they speak to a device

Ghayth-Bouzayeni/tunisian-voice-pipeline

Domain:

Record type:

Visit

Tasks