Arabic-STT/common-voice-18-arabic

Dataset Card for Common Voice 18 – Arabic Edition

Dataset Summary

This dataset is an unofficial Arabic-only extraction of Mozilla Common Voice Corpus 18.0, prepared for Automatic Speech Recognition (ASR) research and development.

It is derived from the original Common Voice 18 release and filtered to include Arabic (ar) speech data only, while preserving the original dataset structure, splits, and metadata fields. The dataset consists of validated, unvalidated, and invalidated speech recordings paired with Arabic transcriptions and speaker metadata.

• Language: Arabic (ar)
• Task: Automatic Speech Recognition (ASR)
• Audio Sampling Rate: 48 kHz
• License: CC-0 (Public Domain)
• Source: Mozilla Common Voice Corpus 18.0

---

Supported Tasks and Leaderboards

• Automatic Speech Recognition (ASR)
• Speech-to-Text (STT)
• Arabic acoustic and language modeling
• Speaker and accent analysis (limited metadata)

---

Dataset Structure

Data Instances

Each data sample includes:

• An audio recording of spoken Arabic
• A corresponding sentence (transcription)
• Speaker metadata (age, gender, accent, locale)
• Voting-based validation information

Features

Feature Name	Type	Description
client_id	string	Anonymous speaker identifier
path	string	Relative path to the audio file
audio	audio (48kHz)	Speech waveform
sentence	string	Arabic transcription
up_votes	int64	Number of positive validations
down_votes	int64	Number of negative validations
age	string	Speaker age group
gender	string	Speaker gender
accent	string	Speaker accent (if provided)
locale	string	Locale information
segment	string	Recording segment ID
variant	string	Sentence variant

---

Dataset Splits

Split	Examples	Size
Train	28,410	~766 MB
Validation	10,471	~313 MB
Test	10,471	~324 MB
Other	41,586	~1.28 GB
Invalidated	15,120	~497 MB
Total	105,058	~3.18 GB

---

How to Use

Load with 🤗 Datasets

from datasets import load_dataset

dataset = load_dataset("MohamedRashad/common-voice-18-arabic", split="train")

Streaming Mode

from datasets import load_dataset

dataset = load_dataset(
    "MohamedRashad/common-voice-18-arabic",
    split="train",
    streaming=True
)

print(next(iter(dataset)))

PyTorch DataLoader

from datasets import load_dataset
from torch.utils.data import DataLoader

dataset = load_dataset("MohamedRashad/common-voice-18-arabic", split="train")
dataloader = DataLoader(dataset, batch_size=32, shuffle=True)

---

Dataset Creation

This dataset was created by:

1. Downloading Mozilla Common Voice Corpus 18.0
2. Filtering the corpus to include Arabic language data only
3. Preserving original splits and metadata
4. Converting the data into Hugging Face datasets format

No additional annotation or modification of audio or text was performed.

---

Licensing Information

This dataset is released under the Creative Commons Zero (CC-0) license, the same as the original Mozilla Common Voice dataset. You are free to use, modify, and redistribute it without restriction.

---

Citation

If you use this dataset, please cite the original Common Voice paper:

@inproceedings{commonvoice:2020,
  author = {Ardila, R. and Branson, M. and Davis, K. and Henretty, M. and Kohler, M. and Meyer, J. and Morais, R. and Saunders, L. and Tyers, F. M. and Weber, G.},
  title = {Common Voice: A Massively-Multilingual Speech Corpus},
  booktitle = {Proceedings of the 12th Conference on Language Resources and Evaluation (LREC 2020)},
  pages = {4211--4215},
  year = 2020
}

---

Acknowledgements

Thanks to Mozilla and the thousands of volunteer contributors who made the Common Voice project possible.