Egyptian and Nigerian English join the Open ASR Leaderboard
Web page · https://huggingface.co/blog/open-asr-leaderboard-liva
First seen by Oriel on .
AI summary
The Open ASR Leaderboard adds private conversational English speech data for Egyptian and Nigerian accents, expanding accent-specific evaluation and reporting.
- Liva AI Conversational EG contains 1.34 hours of Egyptian-accented speech, while Liva AI Conversational NG contains 1.55 hours of Nigerian-accented speech.
- Both datasets contain spontaneous conversational speech with punctuated, cased transcriptions that preserve disfluencies.
- The leaderboard now reports US versus non-US accents and Inner Circle versus Outer and Expanding Circles using Kachru’s framework.
- Liva AI collects multimodal interaction data through consumer products, providing natural, high-context behavior from real product use.
- The private splits are intended to improve representation and reduce benchmaxxing in ASR evaluations through held-out data and additional providers.
#automatic speech recognition #accent evaluation #speech datasets #liva ai
Generated by gpt-5.6-sol. Check the original for details.
Aliases
| Source | External ID | URL |
|---|---|---|
| Hugging Face Blog | https://huggingface.co/blog/open-asr-leaderboard-liva | https://huggingface.co/blog/open-asr-leaderboard-liva Read discussion |