2026
Towards Privacy-Preserving Fine-Tuning: Anonymization of Aphasic Speech for Effective ASR
Fine-tuning on anonymized aphasic speech data improves ASR performance by 18 % compared to the performance of generic models on non-anonymized speech, and this gain in utility is achieved alongside substantial privacy protection.
Sebastian Hofstetter, Timo Baumann
· International Conference on... · 0 citations