Learning to play a musical instrument such as a piano requires many hours of exercises, generally either suggested to the student by a teacher or taken from a “method” book. These books are collections of progressive exercises meant to teach specific techniques and address the commonest mistakes and difficulties that players face while learning. One downside of these books is that the exercises are not personalized to the students, and thus cannot address specific difficulties and characteristics of each learner. Teachers could potentially write or find exercises that are tailored to the different needs of their students, but this activity is very time-consuming, and given the number of students that many teachers have, it becomes prohibitive.

Given the many recent advances in the field of music generation, we propose that it should be possible to generate exercises automatically to form a personalized method for each student. The teacher would describe the characteristics of the student and their strength and weaknesses to a software system, as well as the teaching goals that should be covered in the generated exercises. In return, the system would create a sequence of exercises that are specific to the needs of the student and to the concerns of the teacher. This “interactive opus” would allow for a more effective and engaging learning experience, since the learning difficulty curve would follow each student rather than enforcing a general curriculum.

CAL:IOPE (Co-creativity And Learning: Interactive Opus generation for Piano Education) is a project funded by the European Commission under the MSCA-2022-PF action. With this fellowship we propose to study both a theoretical model for the description of piano learning students, and a computational system to generate progressive exercises based on that model. Advancing knowledge in Computational Creativity and music generation to tackle long-term structure, this project will aim to create an entire set of musical exercises that are coherent in style and difficulty, leading research towards music generation that is useful to human musicians.

In order to achieve these goals, we will both use bottom-up approaches applying machine learning to a dataset of piano exercises we will construct for this project, and a top-down collaborative design approach by interviewing a focus group of experiences piano teachers to find out common difficulties and potential use cases for the system that keep the human teacher in the loop.