Doctoral Colloquium (Music): Yaolong Ju, PhD candidate in Music Technology

Event

Price: 
Free

The Doctoral Colloquium is open to all.

"Automatic Figured Bass Annotation and Automatic Chord Labelling Using the New Bach Chorales Figured Bass Dataset"

Yaolong Ju (music technology PhD candidate)

Friday, 27 November 2020 at 16:30

https://mcgill.zoom.us/j/93441829030

Abstract: This research focuses on the computational study of figured bass, which remains an under-researched topic in music information retrieval research, likely due to a lack of machine-readable datasets. First, we introduce the Bach Chorales Figured Bass dataset (BCFB), a collection of 139 chorales composed by Johann Sebastian Bach that includes both the original music and figured bass annotations encoded in MusicXML, **kern, and MEI formats. We also present a comparative study of automatic figured bass annotation using both rule-based and machine learning approaches, which respectively achieved classification accuracies of 85.3% and 85.9% on BCFB. There are several promising areas for computational music research involving figured bass. One is automatic chord labelling, where figured bass can potentially offer indications of harmonic rhythm and non-chord tones, thereby reducing chord labelling ambiguity. We therefore propose a series of four rule-based algorithms that automatically generate chord labels for homorhythmic Baroque chorales based on both figured bass annotations and the musical surface, and apply them to the new BCFB dataset. The resulting chord label annotations produced by our system are presented as the new Bach Chorales Multiple Chord Labels (BCMCL) dataset, which provides a choice of four parallel chord labels for each of the 139 chorales. These range from one set of labels based only on the figured bass alone, which do not assume any music-theoretical ideas proposed after the time the chorales were written, to a set of labels based on both the figured bass annotations and the full musical surface, and which considers the music from the perspective of modern tonal music theory. It is hoped that this dataset and the algorithms used to label it will be of interest for both future musicological research and research on automatic chord labelling systems.

 

 

 

Back to top