Tools for Listening to Texts-in-Performance

Audio archives provide tremendous resources for studying texts-in-performance, performance styles, and media history and formats. Such research requires tools that work well on low-quality, noisy audio common in humanities research, e.g., poetry readings, radio plays, and talking books, the datasets for this project. The proposed project will develop, provide access to, and support humanistic research using two state-of-the-art, open-source, user-friendly tools, Gentle and Drift. Drawing on advanced speech recognition and signal processing algorithms, Gentle and Drift visualize and quantify prosodic, expressive features of speech, including pitch range, intonation patterns, intensity, and rhythm. The project will also train a network of scholars in using these tools, and solicit and apply their feedback to develop new features to fit their needs. In so doing, the project will provide practical tools, broaden the community of users and develop new digital humanities research on sound.