Introducing Automatic Grading to Music Education

Online music education has seen a rapid expansion with the proliferation of the internet, making it more accessible to people around the world. However, private music lessons can still be expensive, leading many individuals to rely on one-way instructional materials that provide information but do not offer any feedback.

In a recent research paper (Käo, K., & Niitsoo, M. MatchMySound: Introducing Feedback to Online Music Education. In New Horizons in Web Based Learning. Springer International Publishing.), we described and implemented a system that aims to address this problem by adding interactive exercises to web-based self-study materials. This system, called MatchMySound, is a web application that compares two audio files – a pre-recorded example provided by the teacher and one recorded by the student through the application.

The algorithm evaluates the differences between the two audio files in two main dimensions: sound and timing. Sound characteristics such as pitch, intonation, and articulation are analyzed, as well as timing elements such as rhythm, tempo, and overall speed. The feedback is presented to the student as two numerical scores along with two graphs that provide a finer level of detail.

The visualization of the comparison

To evaluate the effectiveness of MatchMySound, we conducted a study involving 108 beginner hobbyist guitarists and two conservatory-level teachers. The guitarists attempted 216 exercises, and the feedback produced by the system was compared to the grades given by the human teachers. The results showed that the system provides strong promise, as the grades it gave corresponded quite well with the grades given by the human teachers.

There was a strong linear correspondence between the scores given by the human teachers and those produced by the algorithm (r=0.68, p < 1e-16), indicating that the algorithm’s scores align well with those of human teachers. When looking separately at the different dimensions, the correlations were slightly weaker (r=0.39 for sound/notes and r=0.65 for timing).

Overall, this system has the potential to improve the quality of online music education by providing interactive exercises and immediate feedback to students. It could also potentially reduce the cost of private music lessons and make them more accessible to a wider range of individuals. However, when asked if they preferred automatic feedback over human feedback, 84% of the participating guitar students answered “No,” while 10% answered “Yes.”

Leave a Reply

Your email address will not be published. Required fields are marked *