Similarity in computational music: a musicologist's approach

The authors examine a number of methods for text processing, principally coming from computational biology, and examine in which manner they can apply to musical analysis. Then, we propose a number of modifications that can be made to these methods to allow for a better application to musical analysis. To this end, we first examine the practices of musical analysis. We focus on the field of extraction of motives. A common approach to this problem is to consider repetitions; whenever some part of the musical text is repeated, it can be considered as a motive. Detecting motives can either be based on a perfect match, or on inexact matching. To this end, the concept of similarity is introduced and analysed, and its meaning is defined in the scope of musical analysis. We also deal with the problem of the representation (or encoding) of the musical text. The role of encoding and its consequences on the application of algorithms is investigated.