Paula Andrea Perez-Toro

Fine-tuning Whisper models for multilingual medical ASR reveals that the best performance hinges on the adaptation strategy, with surprising shifts in internal representations based on language context.

Layer selection for speech-based PD detection is more about the dataset than the model architecture, revealing a critical flaw in current approaches.

SSL embeddings, typically superior for speech tasks, surprisingly lose their edge to hand-crafted acoustic features when classifying mild cognitive impairment from speech, challenging assumptions about representation learning in clinical applications.