A big part of my time at Kadenze as part of the R&D team was to come up with ways to grade and visualize audio that students were submitting for their online coursework.
For many of the online music creation classes we would have the students submit an audio file to be automatically graded. Using Music Information Retrieval techniques along with some Machine Learning, we developed methods to automatically grade a student's audio submission. We also developed tools to visualize their assignments and provide them with informative feedback.
Using concepts found in Albert Bregman's Auditory Scene Analysis, we developed methods to inform the student of the compositional structure of their assignment.
This work was inspired by Martin Wattenberg's work using MIDI files to generate compositional visualizations.
More can be read about this on the Kadenze Blog.
