Skip to content

Dataset · echonova-ai

Arabic and 8 dialects speech recognition benchmark

echonova-ai/arabic-dialect-asr-benchmark

20 hours of annotated Arabic speech in Modern Standard Arabic and eight regional dialects for evaluating ASR systems.

What's inside

A public test subset from the EchoNova Sada paper includes 8,925 audio clips up to 30 seconds long (19.95 hours total) with manual text transcriptions. The data covers Modern Standard Arabic and eight dialects: Egyptian, Emirati, Saudi, Kuwaiti, Qatari, Omani, Syrian, and Lebanese.

What you can do with it

  • Evaluate ASR model quality on various Arabic dialects
  • Test the robustness of speech recognition systems to regional accents
Tasks
Automatic speech recognition
Languages
Arabic
Size
1K–10K rows
Downloads this month
275
Likes
10
Access
on request on Hugging Face

How to load it

from datasets import load_dataset

ds = load_dataset("echonova-ai/arabic-dialect-asr-benchmark")