School of Music
Collaboration: At the Piano and the Research Lab
Jocelyn Dueck is a professional listener. As a collaborative pianist and vocal coach at the Carnegie Mellon University School of Music, she is a specialist on the production of sound and teaches voice majors Italian, French, German and English diction and the nuances of singing that are controlled "from the chin up." She is now also helping researchers develop machine learning tools that can improve how people everywhere read and comprehend written words.
For Professor Dueck, working in the artificial intelligence (AI) space is just one more step in a long career in music and research. She grew up in Canada with musical parents of a choral background, who emphasized attention to detail in vocal music.
"My parents taught us curiosity," Dueck recalled, "and they’d quibble about grammar and pronunciation." With degrees in piano performance and coaching from the University of Minnesota, Dueck went on to serve on the faculty at Bard College, Mannes College, Manhattan School of Music and The Juilliard School before joining the CMU School of Music as an assistant professor of collaborative piano.
Her current work is with the Virtual Readability Lab, a collaboration between Adobe, the nonprofit organization Readability Matters, the University of Central Florida (UCF) and Google, aimed at developing technologies to improve reading proficiency for people from all backgrounds around the world. The Virtual Readability Lab is developing AI to make font, text size, spacing and letter shape recommendations that can be personalized to an individual to make documents more readable.
Dueck is an expert in prosody, "the music of language," as she puts it. Prosody, which is measured by attributes such as volume, duration, pitch and rhythm, is the process by which people place stress and inflection in both music and spoken language that makes it understandable and not disjointed. She created a rubric with criteria that she uses to listen to a study participant reading out loud and to score how fluent the reader is in the language. This data and the tools the team is building will improve digital reading for next-generation reading platforms.
The readability technology could one day be used for many applications, including improving highway signage to reduce accidents, and helping medical and military professionals process text more quickly, and to efficiently identify the most important information in a document.
In addition to the prosody work, researchers are also use eyetracking technology to follow readers’ eye movements to assess how people tend to scan pages of text, which is often not a straightforward line-by-line reading. Dueck is interested in utilizing technology like this to improve teaching and pedagogy for collaborative pianists and other musicians, thus bringing the music-science collaboration full circle. "All of the tools we use for fluency in reading words can be used to enhance fluency in reading music notation," she said.
"Professor Dueck's work with AI serves as evidence that scientific progress relies on the elusive aspects of the human experience, which are at the heart of music making and art, in general," said Jonathan Bailey Holland, the Jack G. Buncher Head of the School of Music. "She represents a growing field of inquiry and research that Carnegie Mellon is well-suited to foster and support."
These technological advances came about through interdisciplinary collaboration between scientists and artists.
"Artists use their imagination every day to create something out of nothing," Dueck said. "And if we all expanded that sense of curiosity beyond the artistic sphere, we would see that the possibilities for interaction between music and other fields are endless."