---
license: apache-2.0
language:
- ar
pretty_name: SA2Dataset
size_categories:
- n<1K
---

## Overview

This dataset contains spoken Arabic digits from 40 speakers from multiple Arab communities and local dialects. It is augmented using various techniques to increase the size of the dataset and improve its diversity. The recordings went through a number of pre-processors to evaluate and process the sound quality using Audacity app.

## Dataset Creation

The dataset was created by collecting recordings of the digits 0-9 from 40 speakers from different Arab communities and local dialects. The recordings were collected using a high-quality microphone in a quiet environment.

The dataset was then augmented using the following techniques:

* **Speed perturbation:** The speed of the recordings was randomly varied by up to 10%.
* **Pitch perturbation:** The pitch of the recordings was randomly varied by up to 10%.
* **Noise addition:** White noise was added to the recordings at a low level.

This augmentation resulted in a total of 3,216 recordings.

## Dataset Statistics

The dataset contains the following statistics:

* Number of speakers: 40
* Number of dialects: 5
* Number of recordings: 402 
* Average recording length: 1 second

## Dataset Split

The dataset is split into the following subsets:

* Train: 282 recordings
* Validation: 80 recordings
* Test: 40 recordings

## Dataset Usage

This dataset can be used for a variety of tasks, including:

* Automatic speech recognition (ASR)
* Speaker identification
* Language modeling
* Dialect identification

## Dataset License

This dataset is licensed under the apache-2.0.

## Dataset Citation

If you use this dataset in your research, please cite the following paper:


... soon ....