A Toolbox for Analyzing non-Western Music

MORTY (MOde Recognition and Tonic I(Y)dentification) is a free-software Python library and can be found here. It implements a handful of domain-specialized machine learning techniques for achieving two fundamental analysis tasks on a given music audio file; classifying its mode (e.g. major & minor in Western music) and/or estimating its tonic (e.g. "C" in a Western song in C-major). These features are often as inputs to more advanced tasks such as tuning, intonation and melody progression analyses.

MORTY was developed during the summer of 2014 in Barcelona, in collaboration with Universitate Pompeu Fabra's Music Technology Group as part of the EU funded CompMusic project (2011-2017). This project aimed to develop computational methods and analyses for non-Western music traditions, with a focus on five traditions; Hindustani (North India), Carnatic (South India), Ottoman-Turkish makam (Turkey), Arab-Andalusian (Maghreb), and Beijing Opera (China).

Given the focus of CompMusic project, MORTY was designed to be culture-agnostic and can be trained on any music tradition. We have used this tool in Ottoman-Turkish makam, Hindustani, Carnatic and Beijing opera with state-of-the-art results. I have later had North African scholars reach out about using it in their local music traditions, as well. More details on its performance can be found in my first publication. As part of this work, we have also curated and published the largest Ottoman-Turkish music corpus at the time, along with the scripts used in the experiments presented in the paper.