MAGOR speech to text transcription

• Main Author: Lim, Yao Xian
• Other Authors: Chng Eng Siong
• Format: Final Year Project
• Language: English
• Published: Nanyang Technological University 2024
• Subjects: Computer and Information Science
• Online Access: https://hdl.handle.net/10356/181070
• Institution: Nanyang Technological University

This project presents the development and deployment of MAGOR, a web application designed to facilitate audio and video transcription services without relying on internet connectivity. This project introduces a locally deployed speech-to-text system that mitigates security risks commonly associated with cloud-based solutions. By leveraging Docker for containerization, MAGOR ensures secure, isolated processing of audio and video data, which is crucial for environments with strict data privacy requirements. The core objective was to dockerize MAGOR and the associated Automatic Speech Recognition (ASR) gateway, ensuring seamless integration between them. This allows users to upload video files via MAGOR, which are then processed by the local ASR gateway to perform speech-to-text translations. The system is designed to support transcription differentiation for up to eight speakers, with distinct colours used to identify each speaker in the transcription. The application leverages modern technologies including React and Node.js to ensure a responsive and efficient user experience. Significant efforts were made to enhance the reliability of the system, including the successful implementation of the ASR request tracking feature, which indicates whether a recording has been successfully processed by the ASR gateway. Additionally, the statistics tab was developed to provide comprehensive insights into the usage and performance metrics of the system, demonstrating the project’s success in achieving its goals. Finally, important maintenance tasks, such as bug fixes, are also covered in this project.