Improvisation as Poetic Computation

Parsons code is apt for studying and comparing the internal structure of songs, i.e. how sound is organized to communicate a thought. Choose a key and encode a tone with the symbol * =start, U=Up, D=Down, or R=Repeat in that key to describe the internal structure of that song, i.e. how sound should be organized to communicate the concept (=song) encoded by the melodic contour described using those symbols. From this, a process can be developed whereby the act of organizing sound provides a model for how people organize around the structure encoded by, i.e. the mood of, that song across contexts.

For resources on the tradition of treating music/song as archival texts:

Denys Parsons, The Directory of Tunes and Musical Themes, S. Brown, 1975.

Henry Louis Gates Jr., The Signifying Monkey, Oxford, 1988. Sections 2.4-2.5.

LeRoi Jones (Amiri Baraka), Blues People, Grove Press, 1963.

—————————- Black Music, William Morrow & Co., 1968

References:

James Weldon Johnson, The Books of American Negro Spirituals, New York: Viking Press, 1925.

Peter Burkimsher, “Chord Progressions,” Github, 2017. (https://peterburk.github.io/chordProgressions/index.html)

See also: https://www.musipedia.org/

Notes:

James Weldon Johnson’s (JWJ) songbook was transcribed into Parsons code using MuseScore Parsons Code Exporter 1.0 from .mxml scores of the songs collected in the JWJ volume. This was done to maintain consistency throughout the translation process. Singing books, in which music was treated as a textual archive, was produced by a machine trained on the Burkimsher and JWJ datasets.

“Songbooks,” i.e. singing texts, were generated using DeepAI’s GPT-3 text generator. To model improvisation, after training our text generator we used as input an encoding of the “kernel” phrase James Weldon Johnson identified in the introduction to his book. This kernel was determined as the generative core of the music emanating from the Americas after the introduction of slaves to both continents and the Caribbean and found in the work songs, spirituals, and eventually the blues. From this core, and modeled by the GPT, we simulated the evolution of blues from spirituals, to jazz, and other genres: Rhythm and Blues, Rock and Roll, etc. This kernel phrase is the following:

*UDRRUDRR or *RRRRRUD

Locations and Playlists of songs played at BLM protests globally can be found here:

http://vpii.us/blm

Acknowledgments:

I would like to acknowledge both the Johannesburg Institute for Advanced Study and the Institute for Advanced Study at the University of Amsterdam and the Humanities at the University of Edinburgh for their support during the completion of this project.

Victor Peterson II contact: petersonv [at] gmail.com