DEVELOPMENT OF A SPEECH TO TEXT TRANSCRIPTION SYSTEM USING NATURAL LANGUAGE PROCESSING (NLP)

Abstract This research focuses on the design and implementation of a speech-to-text transcription system using Natural Language Processing (NLP). The study addresses the limitations of manual transcription, which is often time-consuming, costly, and error-prone, by developing a web-based application that automates the process. The system was built with a client-server architecture: the frontend hosted on Github pages for accessibility and the backend deployed on Hugging face spaces, integrating OpenAI’s Whisper model for transcription, key NLP techniques such as tokenization, text normalization, and punctuation restoration were applied to enhance accuracy and readability of transcriptions. System evaluation showed high performance with clear audio inputs, though accuracy reduced slightly in noisy environments. The findings demonstrate that open-source tools and cloud-based deployment can provide an efficient, cost-effective, and user-friendly transcription solution. This project highlights the potential of NLP-driven speech recognition for applications in education, accessibility, business communication, and digital documentation, while recommending future improvements in noise handling, multi lingual support, and real-time streaming capabilities.