Towards Modelling Variation in Music as Foundation for Similarity

This paper investigates the concept of variation in music from the perspective of music similarity. Music similarity is a central concept in Music Information Retrieval (MIR), however there exists no comprehensive approach to music similarity yet. As a consequence, MIR faces the challenge on how to relate musical features to the experience of similarity by listeners. Musicologists and studies in music cognition have argued that variation in music leads to the experience of similarity. In this paper we review the concept of variation from three different research strands: MIR, Musicology, and Cognitive Science. We show that all of these disciplines have contributed insights to the study of variation that are important for modelling variation as a foundation for similarity. We introduce research steps that need to be taken to model variation as a base for music similarity estimation within a computational approach.