Egyptian and Nigerian English join the Open ASR Leaderboard
The Open ASR Leaderboard has expanded its private evaluation data by adding two new accents—Egyptian English and Nigerian English—through a collaboration with Liva AI, complementing earlier private splits including Indian-accented English. The leaderboard now reports accent breakdowns using Kachru’s Three Circles, distinguishing US vs. non‑US and Inner Circle vs. Outer & Expanding Circles. This richer, private‑data‑driven evaluation aims to improve representation and trustworthiness of ASR benchmarks. Engineers can leverage these more diverse held‑out sets to better gauge model performance across under‑represented accents.
⚡ Key Takeaways
- Added private splits for Egyptian 🇪🇬 and Nigerian 🇳🇬 English accents.
- Accent reporting now follows Kachru’s Three Circles (US vs. non‑US; Inner Circle vs. Outer & Expanding Circles).
- Private data reduces “bench‑maxxing” by limiting over‑fitting to public test sets.
- Data sourced from Liva AI, which collects multimodal interaction data from real consumer product use.
- The leaderboard will further split the Outer and Expanding Circles as more accents are added.
🔧 Tools & Libraries
More diverse, private held‑out data lets engineers validate ASR models on real‑world, under‑represented speech patterns, leading to more robust deployments across global user bases.
✅ Practical Steps
- Apply the concepts from this article to your own system design.
Want the full story? Read the original article.
Read on Hugging Face Blog ↗