Speech Metadata Recorder: Enhancing Data Accessibility and Processing for Speech AI Applications

ABSTRACT The Speech Metadata Recorder is a structured metadata management system designed to enhance speech dataset organization for speech recognition research and AI-based voice analysis. Unlike traditional speech corpora that lack dynamic metadata handling, this system enables real-time metadata entry, auto-generated speaker IDs, advanced search and filtering, data export, and visualization capabilities. The system efficiently processes 500+ metadata records, capturing essential attributes such as speaker demographics, linguistic diversity, recording conditions, and device variations. Through optimized database management and visualization tools, the Speech Metadata Recorder improves data accessibility, usability, and AI model training. Experimental results demonstrate that the system achieves fast metadata entry (under 5 seconds per record), optimized search functionality (retrieval time under 1 second), and seamless export in CSV, JSON, and Excel formats. Future enhancements include AI-driven metadata validation, noise- level detection, and speech-to-text integration for automated annotation. The system serves as a scalable and efficient solution for speech dataset management, contributing to multilingual and robust speech processing