Egyptian and Nigerian English Added to the Open ASR Leaderboard
Egyptian and Nigerian English Added to the Open ASR Leaderboard The Open ASR Leaderboard has added two private evaluation splits covering Egyptian and Nigerian English through a...
By AI Engineering Team
Egyptian and Nigerian English Added to the Open ASR Leaderboard
The Open ASR Leaderboard has added two private evaluation splits covering Egyptian and Nigerian English through a collaboration with Liva AI.
Earlier, the leaderboard added 11 private splits from Appen and DataOceanAI. Those datasets covered scripted and conversational speech across five accents. The project later added a private split for Indian-accented English through Voice Arena. Held-out private datasets help make speech-recognition evaluations more trustworthy by limiting the risk of optimizing specifically for benchmark data.
New private evaluation data
The two new Liva AI splits contain spontaneous, conversational speech:
| Dataset | Accent | Duration (hours) | Style | Transcription |
|---|---|---|---|---|
| Liva AI Conversational EG | Egyptian | 1.34 | Conversational, spontaneous | Punctuated, cased, with disfluencies |
| Liva AI Conversational NG | Nigerian | 1.55 | Conversational, spontaneous | Punctuated, cased, with disfluencies |
The additions support broader representation in automatic speech recognition evaluations by expanding coverage to more accents and, over time, more languages. They also introduce additional private data providers, helping reduce the risk of benchmark-specific optimization.
Expanded accent breakdown in the Private tab
The new accents allow the Open ASR Leaderboard to provide a more detailed accent breakdown based on Kachru’s Three Circles framework. The approach was also used by Hume AI in its leaderboard.
The Private tab now reports results using two groupings:
- US versus non-US: US, UK, Australia, Canada, India, Nigeria, and Egypt
- Inner Circle versus Outer and Expanding Circles:
- Inner Circle: US, UK, Australia, and Canada
- Outer and Expanding Circles: India, Nigeria, and Egypt
As more accents are added, the Outer and Expanding Circles will be divided further.
About Liva AI
Liva AI collects multimodal interaction data through consumer products used by real people. It controls both the data-collection environment and distribution, allowing it to develop products around specific model gaps across domains, industries, languages, modalities, and interaction types. This provides access to natural, high-context behavior that occurs during real product use rather than through artificially elicited or task-based interactions.