Layer-Wise Analysis of Multilingual Medical ASR Adaptation in Whisper
August 20, 2026
Layer-wise encoder analysis shows that direct EN+DE fine-tuning on Whisper-Medium yields the lowest combined English and German word error rates (26.30%). The study examines how multilingual medical adaptation reshapes internal model representations beyond standard WER metrics.
HOW THIS AFFECTS YOU
●
researcherYou can use these findings to select optimal fine-tuning strategies for multilingual medical ASR.
●
healthThis provides insights into the reliability of Whisper-based speech recognition for clinical diagnostic use.