Data structures for music encoding: tables, trees, and graphs

Abstract

One of the challenges in planning a digital edition is the selection of the encoding system and representation format that the edition will use to encode musical information. This decision affects the form of the edition not only through the descriptive capabilities of the encoding format, but also through the format's adherence to certain data structures such as tables, trees, and graphs. Far from being straightforward containers, each of these structures possess unique qualities that constrain the encoding process into considering the music within the parameters of that data model. Data that does not fit must be cajoled into the chosen model, with varying degrees of success. Recent work has developed arguments for the use of certain formats in specific cases based on features and interoperability, but a thorough review of the suitability and sustainability of the underlying data structures has yet to be conducted. This paper explores the problem of music representation in data structures from the perspective of musical domain. By encoding the same passages of music in multiple hypothetical formats it demonstrates the musical aspects and, particularly, the relationships that the data structures commonly used in music encoding variously excel at, privilege, or struggle with representing. I argue that encoding projects should consider the constraints that the data structures of a format impose on their encoding; whether there are instances that would be unsatisfactorily modelled, the data biases this creates, and therefore whether it would be advisable to extend the representation format to accommodate another data structure.