Dataset · oddadmix
3000+ hours of speech across 13 Arabic dialects for ASR
oddadmix/dialectal-arabic-lahgtna-v2
A large corpus of audio recordings and transcripts in 13 Arabic dialects for training speech recognition systems.
What's inside
It contains approximately 611,000 utterances (over 3,000 hours of recording) of mono audio at 16 kHz with text transcriptions. The recordings cover 13 Arabic dialects, with each fragment labeled by its specific dialect. Transcriptions use authentic dialectal orthography rather than Modern Standard English; the metadata also indicates English.
What you can do with it
- Training and testing multi-dialectal ASR models for the Arabic language
- Developing voice-based systems to recognize specific Arabic dialects
- Tasks
- Automatic speech recognition
- Languages
- Arabic, English
- Size
- 100K–1M rows
- Downloads this month
- 4.3K
- Likes
- 33
How to load it
from datasets import load_dataset
ds = load_dataset("oddadmix/dialectal-arabic-lahgtna-v2")