Abstract:
Automatic Speech Recognition has been a prominent sector in Computer Science Research for decades, generating thousands of research papers in recent years. It is a complex and evolving field, having intersections with ML, NLP, DL and other prominent AI sectors. Being a Complex (involving many steps), Diverse (lots of ways to implement each step, also differing according to the final task) and Computationally heavy field, it had a relatively smaller practitioner base. With the revolution in the Chip Industry, the problem of computation has been solved. The only problem remains in reducing the complexity so that even amateur computer professionals can start their journey on ASR and increase their depth gradually. SpeechBrain, released in 2019, is the exact solution to that problem. It is an all-in-one and user-friendly toolkit that can be used to learn and develop state-of-the-art speech systems aimed at different Speech-related problems. In this report, I have included chapters that are necessary for having a basic understanding of ASR, a basic knowledge of SpeechBrain Repository, and finally, how I have worked in and around this Repository, changed architectures & developed an interactive Web Application capable of Text Generation & Automatic Speech Recognition using SpeechBrain.