Poems through Poems

With computer poetry, the output is always constrained by the input.  The haiku written by Poesytron is limited by the material I've given it to work with.  First, I gave it almost complete free rein: words were selected entirely at random from the Moby hyphenation lexicon.  Then I constrained it to words from a sample of 500 haiku, with some rules on selecting words based on them appearing in the same haiku together.

I have ideas for expanding the sample, and also creating better rules for word selection.  But before I get into those, I want to explore some other ways of applying constraints--by drawing inputs from a particular text.

For example, I can tell Poesytron to create a random-word haiku, but only using words found in Lewis Carroll's Alice in Wonderland:

Tea-time rate removed

Tis putting side sign shyly

lives ever sand tongue

But I could also put constraints on word choice based on how frequently a word appears in a text.  This is easiest to do with something much shorter than a novel; say, Shel Silverstein's poem "Where the Sidewalk Ends":

sidewalk asphalt wind.

Cool grows with grass-- go There slow,

where in, and sun, the

The difference is subtle, but present.

Experimental poets often play with different constraints like these.  I've mentioned Jackson Mac Low, a poet who played with "non-intentional" methods, before, but he had one specific method that drew on other texts in a similar way.  He called it a "diastic" method, and he would read through a source text and replace each single word with several words from another text (the seed text), selecting words in order that had matching letters in certain places.

There's also a group of poets who create "potential literature" through many different methods of constraint.  They're called OuLiPo, and it's mainly a French-speaking group of writers and mathematicians.  (Their website is here; it's in French but Google can translate it, for however much that is worth.)  One of their many methods is called Chimera, and it involves taking one text, replacing all of the nouns with nouns from a second text, all of the verbs with verbs from a third text, and all of the adjectives with adjectives from a fourth text.

I'm intrigued by all of these different methods to "read through" something.  It reminds me of the statistical technique of resampling, which allows a scientist to get a much better understanding of her data by selecting and rearranging the original dataset.  I think these methods of resampling a book or poem--of rearranging, reselecting, and restructuring--can not only create new poems, but can possibly even present new ways of looking at the original text.