ToolsNews
Open ASR Leaderboard adds Egyptian, Nigerian English
Hugging Face's Open ASR Leaderboard adds two private Liva AI conversational sets for Egyptian and Nigerian English accents, widening representation and cutting benchmaxxing risk.
· 1 min read
What changed?
Hugging Face says the Open ASR Leaderboard now includes two new private accent splits from a Liva AI collaboration: Egyptian and Nigerian English.
Both sets are spontaneous conversational speech, transcribed with punctuation, casing, and disfluencies. They sit on the Private tab as held-out data rather than a public downloadable benchmark.
Dataset | Accent | Hours | Style |
|---|---|---|---|
Liva AI Conversational EG | Egyptian | 1.34 | spontaneous conversation |
Liva AI Conversational NG | Nigerian | 1.55 | spontaneous conversation |
Why it matters
This continues earlier private additions from Appen, DataOceanAI, and Voice Arena. The stated goals are broader accent coverage in ASR evals and fresher providers so teams cannot overfit public leaderboard sets.
The Private tab also gets a richer accent breakdown inspired by Kachru's Three Circles: US vs non-US, and Inner Circle (US, UK, Australia, Canada) vs Outer and Expanding Circles (India, Nigeria, Egypt).
For builders picking English ASR models, non-US spontaneous speech now weighs more in the ranking. If you already track Nemotron 3.5 Transcribe on Open ASR, check the new Private columns before you lock a production stack.
Produced with the help of AI, reviewed by an editor. How we make our articles
Reviewed and edited by the Neural Daily desk. About
