Poems generated by discovering midpoints between how phrases sound and are spelled | Boing Boing

Poems generated by finding midpoints between how words sound and are spelled | Boing Boing


Allison Parrish is one in every of my favourite poets; she writes code that transforms language into fascinating and evocative literary forms.

Not too long ago she wrote software program that generates “compass poems”. She’ll plot phrases within the NSEW factors on a compass, and her software program interpolates new phrases within the in-between areas, by discovering midpoints between how the unique phrases are spelled and the way they sound.

BOMB magazine published a few of them here (that picture above is from their web site), and Parrish has a whole chapbook of them here, printed as a part of the sync collection.

One pleasant one is the Beatles arranged in the compass points, from BOMB

Or, from the chapbook, the large Web giants …

Or the cardinal virtues of conventional Christian theology …

Parrish’s work jogs my memory of the concrete poetry of bp nichol, a Canadian artist who — from the 60s to the 80s — equally dove the phonemic muck of language, remixing phrases into trippy, often-illuminating new combos. Like many concrete poets of the time, he played around a lot with typographic layout, utilizing every part from the electrical typewriters to, in the last years of his too-short life, software. It is significantly unhappy to me that he died simply earlier than the net and open-source natural-language-processing instruments emerged; I think about he’d have discovered these to be deliriously enjoyable instruments to play with, in exactly the way in which Parrish makes use of them.

At any price, go learn that whole chapbook of Parrish’s — it’s just terrific. I am gonna print it up; I need to have a look at ’em entombed in atoms, holding them in my arms like an animal.

If you wish to know extra about how her code works, she describes it on the finish of the chapbook …

I skilled a machine studying mannequin with two components: a “speller,” which spells phrases based mostly on how they sound, and a “sounder-out,” which sounds out phrases based mostly on how they’re spelled. Within the strategy of sounding out a phrase, the “sounder-out” produces a fixed-length numerical vector, often known as a “hidden state,” which is basically a condensed illustration of a phrase’s phonetics. The “speller” can then use the phonetic info contained on this hidden state to provide a believable spelling of the phrase. The hidden state, like some other numerical vector, might be modified: translated, multiplied, blurred, averaged.

Every of the poems collected right here outcomes from a pc program I wrote that performs the next steps: (1) use the “sounderout” to search out the hidden state for 4 phrases, drawn from a handauthored listing (these are the phrases on the “factors” of every poem); (2) discover the vector midway between the hidden state vectors for every pair of “level” phrases, and predict a believable spelling for these midway vectors with the “speller”; (3) discover the vector of the midpoint of all eight vectors produced in (1) and (2), and likewise predict a believable spelling for this vector.

The phrases from steps (1) and (2) are then programmatically organized within the type of a compass rose. The phrase ensuing from step (3) is positioned within the center.





Source link

Leave a Reply

Your email address will not be published. Required fields are marked *