A new system called oollim allows for the evaluation of L2 English speech on a consumer gaming PC, focusing on accuracy, fluency, and prosody. This system integrates pretrained representations from XLS-R, XEUS, and PhoneticXeus with reference-text features. Trained on several speech corpora, oollim has demonstrated superior performance compared to the Microsoft Azure Pronunciation Assessment API in correlation and kappa metrics on independently held-out recordings. AI
IMPACT This research could enable more accessible and cost-effective speech analysis tools for language learners and educators.
RANK_REASON The item is an academic paper detailing a new system for speech evaluation. [lever_c_demoted from research: ic=1 ai=1.0]
- J-AESOP
- Korean Learners' L2 Speech Corpus
- Microsoft Azure Pronunciation Assessment API
- oollim
- PhoneticXeus
- Richard Rose
- SpeechOcean762
- XEUS
- XLS-R
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →