Search papers, labs, and topics across Lattice.
This paper introduces Speech Playground, an interactive tool designed to facilitate the visualization and comparison of speech data by integrating a Python backend with a user-friendly web frontend. By supporting various feature types and offering functionalities like TextGrid and forced alignment, it streamlines the process of analyzing speech in conjunction with modern deep learning techniques. The key result is that Speech Playground enhances the usability of speech analysis, making it more accessible for researchers and practitioners in fields such as speech research and computer-aided pronunciation training.
Speech Playground transforms speech analysis by enabling seamless integration of deep learning representations with interactive visualization tools.
This paper presents Speech Playground, an interactive speech visualization and comparison tool. While existing tools such as Praat are excellent, it can be cumbersome to integrate them with modern deep learning representations and use them for comparison. Speech Playground addresses this by combining a Python backend with a web-based frontend for interactive exploration of multiple feature types, including continuous, discrete, and variable-length representations. It includes TextGrid and forced alignment support together with configurable distance and alignment settings for visual and auditory comparison. Speech Playground is intended for use in speech research, representation validation, and computer-aided pronunciation training (CAPT)-oriented experimentation.
