A computational exploration of melodic patterns in Arab-Andalusian music

Abstract

Here we present a computational approach to identifying melodic patterns in a dataset of 145 MusicXML scores with the aim of contributing to centonization theory in the Moroccan tradition of Arab-Andalusian Music – a theory in development by expert performer and researcher of this tradition, Amin Chaachoo. Central to his work is the definition of a set of characteristic patterns, or centos, for each ṭab‘, or melodic mode. We apply three methods: TF-IDF, Maximally General Distinctive Patterns (MGDP) and the Structure Induction Algorithm (SIA) to identify characteristic patterns at the level of ṭab‘. A substantial number of the centos proposed by Chaachoo are identified and new melodic patterns are retrieved. A discussion with Chaachoo about the obtained results promoted the elicitation of other categories of recurrent patterns in the tradition different from the centos, contributing to a deeper musicological knowledge of the tradition.