Semi-supervised learning for speech recognition

• Tech Stack: Python, scikit-learn, numpy, scipy, soundfile, jupyter notebook, pytorch
• Google Drive URL: Project Link

Introducing semi-supervised learning for speech recognition and highlight its advantages compared to supervised learning. Focusing on training speech recognition models with a small set of annotated data and generating pseudo-labels for a larger set of unannotated data. The evaluation metrics employed are test accuracy, Phoneme Error Rate (PER) and edit distance.