Open ASR Leaderboard adds Egyptian and Nigerian English private data via Liva AI

·2026.11.06 09:00

Key point

New private splits from Liva AI include 1.34 hours of Egyptian and 1.55 hours of Nigerian conversational speech to improve evaluation reliability.

1 / 2

Hugging Face and Liva AI added two new private data splits to the Open ASR Leaderboard covering Egyptian and Nigerian English accents.
The datasets consist of 1.34 hours of Egyptian and 1.55 hours of Nigerian conversational, spontaneous speech with punctuated, cased transcriptions including disfluencies.
The leaderboard now reports accent breakdowns based on Kachru’s Three Circles, distinguishing Inner Circle (US, UK, Australia, Canada) from Outer and Expanding Circles (India, Nigeria, Egypt).
Private data is used to minimize benchmaxxing and improve trust in evaluations, following previous additions from Appen, DataOceanAI, and Voice Arena.

This summary was generated automatically by AI. Check the original for the author's claims and context. Copyright belongs to the original author.

Our guide explains how the AI works. Report summary errors, attribution issues, or removal requests via Contact.