One of our founders, Perry R. Cook, is writing a 3-part series about online music education on Class Central. Here’s a snippet of part two: the dawn of Robot TAs.
We at Kadenze make extensive use of our own machine learning algorithms and user/instructor feedback displays to provide assessment and feedback on a large variety of media-rich assignment types, including sound/music, images (photos and hand-drawn pictures), and videos.
This fits particularly well for many music courses, where a sound or music file can be evaluated automatically, marked, and displayed back to the student with indications as to how well they did at executing the required criteria. Figure 1 shows auto-grader feedback for a music production assignment, highlighting differences from a perfect submission.
Figure 1: Robot grader student feedback for production mixing/EQ assignment.
With machine-intelligent DSP-based music accompaniment software, some systems and companies have made strides toward attaining the elusive goal of private music instruction (or at least computer-mediated rehearsal). Roger Dannenberg (of CMU) has done seminal work on automatic music accompaniment systems, yielding SmartMusic, which is licensed to Coda.
These systems were installed in music studios and practice rooms, and could “listen” to the student, accompanying them and noting/correcting mistakes as they were played. SmartMusic was then licensed to MakeMusic! Inc. and sold as downloadable computer software (on a subscription model).
Dannenberg now works with Music Prodigy, Inc. to provide computer/mobile “Quantified Music Practice” applications. These allow students to rehearse their instruments with accompaniment, then transmit a machine-graded performance to their teacher.
