- Performing Big DataAlan Lomax and The Folk-Song Science
At first glance, computers are at odds with live performance. The cold predetermined logic of the algorithm would seem to negate the spontaneity valued by artists and audiences in even the most scripted and rehearsed productions. My purpose, in structuring most of my operas and audiovisual works around computational concepts, has been to heighten this incongruity into a theatrical tension that speaks to a distinctly contemporary condition. On one hand, algorithmic processes have become increasingly internalized, often determining what people think, do, and want in the world; on the other hand, there is still something—call it a classical sense of free will, or perhaps simply the friction of particular bodies and voices—that resists such automation. This was the kernel of previous works such as I/O (2006) and BOTCH (2013), where vocalists interact with each other and with machines according to rules cannibalized from computer programs, customer service protocols, and voice recognition devices.
This same tension between agency and technological constraint led me to Alan Lomax as the subject of my most recent opera, OYSTER. Though he is popularly known for his recordings of blues and folk music in the 1940s and 1950s, the American folklorist and musicologist spent much of his later career on what we today would call a big data project. In the early 1960s, Lomax embarked on Cantometrics, a system he invented to profile and quantify different styles of folk music from every known culture in the world. It is difficult to overstate the ambition of this enormous institutional project, housed by Columbia University and funded by the federal government, that included an interdisciplinary research team of linguists, anthropologists, statisticians, programmers, and movement analysts—all fueled by the general cybernetic fervor of the time. Harnessing the processing power of an IBM mainframe computer, the goal was no less than to capture a computable image of the entire planet. My libretto for OYSTER asks how Lomax, the folk song fellow traveler, could become the engineer of this total [End Page 58] system with imperial dimensions. My vocal score and video design questions whether such a system can make universal claims in the first place.
OYSTER premiered at Roulette Intermedium in New York City, and was staged as a public lecture by Lomax (performed by John Rose). From his podium, he explains his "folk song science" in a sung-spoken style, accompanied by live prepared piano and electronics. His presentation is interspersed with a trio of performers (Christina Campanella, Michael Chinworth, and Saori Tsukada) who piece together Lomax's life, parsing and scanning an extensive, recently declassified FBI file, while performing a choreography of geometric patterns staged by Phil Soltanoff. Tightly woven into the live performance are a series of projected music films that function as audiovisual aids for the lecture, as well as being the musical core of the opera. In each film we see and hear an ensemble of vocalists (the same performers that are live onstage) simulate a song associated with a different region of the world.
Here it will help to explain the basics of Cantometrics. Imagine a room full of musicology graduate students in 1964, wearing headphones plugged into reel-to-reel tape machines. The actual conversion of individual songs into data was a hybrid human-machine process whereby these trained coders would carefully analyze recorded songs, extracted from hundreds of different local cultures throughout the world. Whether it was a lullaby recorded in an Indonesian village or a choral wedding song from the Balkans, various attributes were numerically rated. Tempo, for example, was rated on a scale from one to five: a quickly paced tune may be rated a four whereas a funeral dirge might be a one or two. Other attributes, many of which had not previously been considered quantifiable, included the degree of vocal harshness, the enunciation of consonants, or how well a group of singers blended together. For each song exactly thirty-seven such attributes were rated, making up its profile.
The contemporary resonance is that Cantometrics was not concerned with individual songs, but with the statistical analysis of thousands of songs...