Pandoo: ASR for Foreign Language Learning
Overview
This Automatic Speech Recognition (ASR) system is built to assess Bengali pronunciation for non-native speakers. Through a novel Phonetic Distance metric, it provides accurate pronunciation scoring, assisting learners in overcoming linguistic barriers and improving their language skills.
Highlights
- Unique Scoring Metric: Introduces Phonetic Distance, which accurately scores pronunciation by addressing character-phone variability and Unicode issues common in Indic languages.
- High-Performance Transformers: Uses advanced transformer models for precise speech recognition, suitable for interactive language learning applications.
Educational Impact
This ASR tool supports language learning for non-native Bengali speakers by providing clear, actionable feedback on pronunciation, aiding both independent learners and structured language programs.