Dual Student is a semi-supervised learning algorithm originally introduced for image classification by Ke et al. In this work, we implement, validate and extend it for Automatic Speech Recognition (ASR). See the report for more details. This project is part of the Speech and Speaker Recognition course (DT2119) at KTH.
pip install dualstudent-asr