README.md

---
license: cc-by-nc-4.0
language:
- ar
multilinguality:
- multilingual
task_categories:
- automatic-speech-recognition
- audio-classification
- text-to-speech
tags:
- arabic
- arabic-dialects
- egyptian-arabic
- tunisian-arabic
- gulf-arabic
- levantine-arabic
- speech-data
- voice-ai
- asr
- tts
- middle-east
- north-africa
- msa
- dialectal-arabic
pretty_name: "🇸🇦 Arabic Dialects Speech Dataset"
size_categories:
- n<1K
---

# 🇸🇦 Arabic Dialects Speech Dataset

<p align="left">
  <img src="https://cdn-uploads.huggingface.co/production/uploads/69162b50b89e7abe20de4b5a/LWhs4p2lPFcyiVsP0tluu.png" width="40%">
</p>

[![Website](https://img.shields.io/badge/Website-silencioai.com-blue?style=flat-square)](https://www.silencioai.com)
[![Contact](https://img.shields.io/badge/Contact-sofia@silencioai.com-green?style=flat-square)](mailto:sofia@silencioai.com)
[![Data Available](https://img.shields.io/badge/Full_Corpus-8,000+_hours-orange?style=flat-square)](mailto:sofia@silencioai.com)

---

> **🌍 Real Arabic dialects from native speakers across the Arab world.**
>
> **📧 Need more?** [sofia@silencioai.com](mailto:sofia@silencioai.com) — we have **8,000+ hours** of Arabic dialect data available.

---

## 🎯 Dataset Overview

**187 recordings** across **4 major Arabic dialect regions**, sourced from native speakers.

| Dialect | Code | Samples | Countries |
|---------|------|---------|-----------|
| 🇪🇬 **Egyptian** | `egyptian` | 117 | Egypt (Cairene, Alexandrian) |
| 🇹🇳 **Tunisian/Maghrebi** | `tunisian` | 55 | Tunisia, Algeria |
| 🇸🇦 **Gulf** | `gulf` | 11 | Saudi Arabia, UAE, Qatar |
| 🇯🇴 **Levantine** | `levantine` | 4 | Jordan, Syria, Lebanon |

## 🚀 Quick Start

```python
from datasets import load_dataset

# Load Egyptian Arabic
egyptian = load_dataset("SilencioNetwork/arabic-dialects-speech", "egyptian")

# Load Gulf Arabic
gulf = load_dataset("SilencioNetwork/arabic-dialects-speech", "gulf")

# Load Tunisian Arabic
tunisian = load_dataset("SilencioNetwork/arabic-dialects-speech", "tunisian")

for sample in egyptian['train']:
    audio = sample['audio']
    transcript = sample['transcript']
    dialect = sample['dialect']
    print(f"[{dialect}] {transcript[:50]}...")
```

## 🌍 Why Arabic Dialects?

Arabic has massive dialectal variation — a model trained on MSA (Modern Standard Arabic) often fails on dialectal speech:

- 🇪🇬 **Egyptian Arabic** — Most widely understood (media/film influence)
- 🇸🇦 **Gulf Arabic** — Business and energy sector
- 🇲🇦 **Maghrebi** — North Africa, French influence
- 🇯🇴 **Levantine** — Syria, Lebanon, Jordan, Palestine

This dataset captures **real dialectal speech** from native speakers.

## 📊 Full Data Availability

| Dialect | Sample | Full Corpus |
|---------|--------|-------------|
| Egyptian | 117 | **3,500+ hours** |
| Gulf (Saudi, UAE, Qatar) | 11 | **1,800+ hours** |
| Maghrebi (Tunisia, Morocco, Algeria) | 55 | **1,200+ hours** |
| Levantine (Jordan, Lebanon, Syria) | 4 | **900+ hours** |
| Iraqi | — | **400+ hours** |
| **Total Arabic** | **187** | **8,000+ hours** |

## ⚖️ License

**CC BY-NC 4.0** — Free for research and non-commercial use.

For **commercial licensing**, contact [sofia@silencioai.com](mailto:sofia@silencioai.com).

## 📧 Get More Data

**📧 Email:** [sofia@silencioai.com](mailto:sofia@silencioai.com)

**🌐 Website:** [silencioai.com](https://www.silencioai.com)

---

## Citation

```bibtex
@dataset{silencio_arabic_dialects_2025,
    title        = {Arabic Dialects Speech Dataset},
    author       = {Silencio Network},
    year         = {2025},
    publisher    = {Hugging Face},
    license      = {CC BY-NC 4.0}
}
```