Meadow9 web application – digital audio recorder for automatic speech recognition

Abstract

This project focuses on enhancing Meadow9, a web application that provides audio transcription services. Previously, Meadow9 relied on manual uploads of local audio files for speech recognition and had not been updated for several years. To identify areas for improvement, a comprehensive analysis of similar products was conducted and compared against Meadow9. This analysis provided valuable insights that informed subsequent use case evaluations and guided the proposal of new functionalities. Additionally, an assessment of Meadow9’s existing technology stack and architecture was carried out to ensure compatibility with the proposed enhancements. Based on these findings, a new architectural design was developed, introducing a dedicated backend system to support the planned improvements. The user interfaces for new frontend components were also carefully designed to accommodate the additional functionalities.

As part of this project, a highly customizable audio recorder with crash recovery was implemented, along with key features such as recorded audio playback, real-time audio visualization during recording and playback, seamless submission of recordings for speech recognition, and robust recording management capabilities. The API was also updated to leverage the latest advancements in the speech recognition engine, while the user interface was refined to enhance usability. The system underwent rigorous testing through both automated and manual methods before being containerized, orchestrated, and deployed to a cloud-based environment. This project successfully addressed the limitations of Meadow9, transforming it into a more user-friendly and feature-rich platform.