Markov Chains and Uncreative Writing
Uncreative Writing and King James Programming
In the Text & Code class which I am taking at RIT for the liberal arts immersion of the same name, we have discussed how digital tools are allowing writers and artists to interact with text in many new and interesting ways. In Uncreative Writing: Managing Language in the Digital Age, Kenneth Goldsmith discusses texts created from other, already existing texts, which he refers to as "uncreative". Some ways in which these texts can be formed include patchworking multiple texts together, selectively reproducing texts, or creating a system to generate text (Goldsmith).
One text that meets Goldsmith's standard for an uncreative text is King James Programming, created by Michael Walker (barrucadu). The "text" as a whole, the blog, is a collection of short text snippets. These snippets are generated using a Markov-chain-based program created by Walker which combines text from different sources in a way that makes the result appear grammatically correct, but which are usually nonsensical and humorous. For King James Programming, Walker trained his program on the King James translation of the Bible, Structure and Interpretation of Computer Programs, and Why's Poignant Guide to Ruby (Walker).
The combination of the archaic, poetic speech of the King James Bible, the conventional academic language of Structure and Interpretation of Computer Programs, and the mad programmer ramblings of Why's Poignant Guide to Ruby create some hilarious results. For example
the appearance of a loop can depend on the order of events in the different cases according to the law of sin.
and
For if ye love them which love you, what reward have ye? do not even mention “classes” or “inheritance.”
and
So, then, one day he was shopping and he found a certain sorcerer, a false prophet, a Jew, whose name was Job; and that man was born in Japan. Which is freaky.
King James Programming meets Goldsmith's standards for uncreative writing because it is made by stitching multiple original texts together to form new texts, and it is created by a system designed to generate text (the Markov chain text generator) rather than a person. One could even argue that it is selectively reproduced, as Walker chooses which generated text snippets make it onto the blog, and which are instead placed on the reject blog.
Experimenting with Markov Chain Text Generation
I decided to generate some text for myself using Markov-chain-based text generation. I went to the Github page for Michael Walker's Markov chain text generator, downloaded the repo as a zip, installed the dependencies, and extracted the zip into a folder in my Documents directory.
I then needed to decide on two or more text sources from which to generate my snippets. This proved to be the most difficult part. I spent several hours searching for public domain texts, and trying different combinations to see if they would generate interesting snippets. Ultimately, I settled on using four poems: Edgar Allen Poe's "The Raven", William Butler Yeats's "The Stolen Child", Robert W. Service's "The Cremation of Sam McGee", and the folk-song "Rynadine (The Mountains High)" for which I constructed a version using my favored phrases from the version presented in full on the Wikipedia page, and from the version on contemplator.com, since no standard, agreed-upon version exists.
The program allows you to specify an integer along with your sources, which appears to define the granularity of your resulting text. I started with 2, which gave me several interesting snippets, but ones that were fairly short in length. Notable examples:
It was lashed to the hateful thing And it harkened with a sudden cry "Is my crematorium!"
She fainted in my arms there all on the marge of Lake LeBarge
Take thy beak from out that shadow that lies floating on the mountains high.
Where dips the rocky highland Of Sleuth Wood in the snow I wrestled with grisly fear
Much I marvelled this ungainly fowl to hear him sizzle so.
Come away, O human child! To the waters and the flames they soared, Such a blaze you seldom see.
The furnace roared and the wind and nothing more.
I also found that at this granularity, the program had a problem with getting caught in loops. Perhaps this is due to the repetition of key phrases in poems, and the granularity defined by 2 happened to be the right length to capture it. For example, "The Raven" got caught in a loop and the text that resulted read
But the Raven, never flitting, still is sitting, still is sitting, still is sitting, still is sitting, still is sitting, still is sitting, still is sitting, still is sitting, still is sitting, still is sitting, still is sitting, still is sitting, still is sitting, still is sitting, still is sitting, still is sitting, still is sitting, still is sitting, still is sitting, still is sitting, still is sitting, still is sitting, still is sitting, still is sitting, still is sitting, still is sitting, still is sitting, still is sitting, still is sitting, still is sitting, still is sitting, still is sitting, still is sitting, still is sitting, still is sitting, still is sitting, still is sitting
I then tried a granularity of 3, but this caused the resulting text to stay within one poem, only stitching together pieces of that poem. I then tried a granularity of 1, but this resulted in several jumps per sentence, creating completely nonsensical gibberish. 2 seemed to be my best bet.
Conclusion
The question that needs to be asked is, "Are texts generated in this way 'creative', or 'uncreative'?" As Goldsmith says of his students in his "Uncreative Writing" class, "what we had accomplished was not uncreative at all; by not being 'creative,' (the student) produced the most creative body of work of her life." (Goldsmith).
We cannot help but be creative in these texts. Each choice shapes the final product. What input texts are used, the granularity chosen, even which snippets I chose to display here as "notable examples". Goldsmith claims "The act of choosing and reframing tells us as much about ourselves as our story about our mother's cancer operation" (Goldsmith). The difference is that the choices are now regarding the process used to create the art, rather than the direct content of the art itself. "The idea becomes a machine that makes the art" (Goldsmith).
Works Cited