Anyone who learns a musical instrument wants to know how is he/she doing. In other words, learners need feedback. Sadly, two of the most popular learning models today – online learning and learning by textbooks – tend to omit it almost completely. But what about the traditional and luxurious one-by-one teaching? While we must admit that this model is all about feedback, it is actually suffering from the same problem. For example, consider the 10 000 hour rule: if a student would practice only in the class and the lessons are twice a week, it would take 96 years to acquire the expert skills. So obviously most of the practicing is still done alone and without professional feedback. One may then reply ‘So what, it’s been like that for a thousand years!’, but the enormous and unprecedented competition that we have today puts musicians under considerably more pressure than ever before. People have less time and, at the same time, thanks to mass media are also too well aware of exactly how good the top players in their game are. That is the driving force behind finding more effective solutions for music education at all levels.
Human and Automated Feedback
Alright then, more and better feedback might help. But where to find such a super patient (while also good yet affordable) music teacher that stands by while a student is practicing and ensures that everything is done correctly? Would anyone want to practice under constant surveillance at all? Experimenting with automated feedback in music education already has some years of history and research suggests that the best results are obtained if human and automatic feedback are combined. There are certain things that humans do better than the algorithms and vice versa. For example, our MatchMySound can be very picky when it comes to technical details: correctness of intonations, length of notes (rhythms, articulations), accuracy and changes of speed. Even after 15 years of teaching and 2000+ students, I’m nowhere close to the level of feedback the algorithm gives. Not in precision or speed, and definitely not in cost.
Sound Recognition vs Comparing to Etalons
We are not the only ones aiming to fix the broken feedback link in music education. Existing solutions using sound recognition tend to compare user input to some form of notation: MIDI, standard sheet music or tablature. We believe that we have improved the situation a great deal as we are comparing the students’ audio input to what their teachers have recorded. On one hand it means that students do not have to watch the screen while playing. The computer is just an assistant: sitting quietly and listening. But most importantly – comparing the audio files of the teacher and student makes it possible to spot possible differences in great detail. Plus, the motivating effect of a teacher as a role model should not be underestimated. Exactly how often did your teacher play your homework fluently in front of you while you were studying?
Teachers Need Feedback, too!
Last but not least, while recording the audio etalons to their students, the teachers also get feedback on their own playing. Maybe even more importantly, the teachers also get feedback about the quality of their exercises. For example, some of their exercises may be completed too easily while the others take too much time to be mastered fluently. For many teachers, it is hard to admit that the exercise they have assigned is too hard for the student, so we just advise practicing more instead of adapting the exercise plan. That’s what I hated most about my own studies. Although I believe that there is always a lot to learn, leaving the classroom and knowing that I still cannot play the instrument although I had practiced 2 years, 6 years, ultimately 20 years, really killed the motivation. That is also where we can see that giving objectively measurable homework can help to motivate music students. Yes, these are just intonations, articulations, rhythms and tempos that we can measure today (with great accuracy, though!) but there will be no art until you have learned your scales and studies!
For those who are accustomed to reading research papers, here are two published research articles by Kristo Käo and Margus Niitsoo: