Generating Music Using Concepts from Schenkerian Analysis and Chord Spaces


Donya Quick


Music demonstrates structure at many levels. This structure can be found in tonalities as well as in temporal aspects of the music. However, many approaches to algorithmic composition neglect the structure at one of these levels, either requiring existing musical fragments as input or not imposing sufficient global structure. The results of these algorithms have a disorganized or random sound. This paper presents a more balanced approach that considers structure at both levels.

Schenkerian analysis is a method for analyzing classical Western music based on recursive simplification of harmonic structure in compositions at different levels of abstraction. This paper describes a generative grammar for classical Western chord progressions using ideas from Schenkerian analysis. The output from the grammar gives the general structure of a chord progression over time, but is still an abstract representation that requires additional information to become a complete score.

Recent research has attempted to find mathematical relationships between harmonies as a way to model perceived similarity between various chords. This kind of model is necessary to reduce the solution space for individual notes to a manageable number of choices. The structural grammar inspired by Schenkerian analysis is combined with these mathematical models for harmony to improve voice-leadings and create a melodic backbone that requires little modification to produce melodies. Results from an implementation of this approach demonstrate a structured sound and simple melodies while avoiding the need for existing music as input.



 title = "Generating Music Using Concepts from Schenkerian Analysis 
          and Chord Spaces",
 author = "Donya Quick",
 institution = "Yale University, Department of Computer Science",
 year = "2010",
 month = "May",
 number = {YALEU/DCS/RR-1440}