OFFICIAL Written by GPN · Nov 7, 2026, 8:30 AM ET · Hugging Face · AI
Hugging Face has added two new accents to its Open ASR Leaderboard: Egyptian English and Nigerian English.
The additions come through a new collaboration with Liva AI, which supplied private conversational speech data for both accents. The Egyptian set runs 1.34 hours and the Nigerian set 1.55 hours, both spontaneous conversation transcribed with punctuation, casing, and disfluencies intact.
It builds on private data added earlier in the year: eleven splits from Appen and DataOceanAI covering scripted and conversational speech across five accents, plus an Indian-accented split from Voice Arena. Private, held-out sets are the point — they keep evaluations trustworthy by making it harder to train against the test data.
The leaderboard now also breaks accents down in more detail, reporting US versus non-US accents and Inner Circle versus Outer and Expanding Circle accents, following Kachru's Three Circles model. The US, UK, Australia, and Canada sit in the Inner Circle; India, Nigeria, and Egypt fall in the Outer and Expanding Circles.
The aim is straightforward: better representation in speech-recognition evaluation, and results the community can trust.
