A Formalist Approach to Computer-Generated Sonnets

Generative Poetry • Durham, NC • Ongoing
cover image

This page will be updated as this project progresses.

I have always felt that my work is connected to poetry, particularly those branches of poetry concerned with the formal arrangement of language on a page. One of my first small games - part of my grad school portfolio - was a look into the way that the sonnet (14 lines, a particular rhyme scheme, a shift in tone after the eighth line) could operate as a structure for a game (14 screens, a 'mechanics rhyme scheme', a shift in tone after the eighth screen). My current work is focused on engaging more deeply with these and other formal structures of poetry in order to produce a computer system for generating sonnets. Specifically (looking forward to the 2018 Creative Turing Tests), the program will be able to generate an infinite number of poems based on an arbitrary noun input.

A common misconception about computers and generative systems is that their intent is to replace artists; this is often not the case. This work is interesting because of the way that poets (particularly me, as a poet) can better understand the mechanics of poetry through the articulation of the formal systems on which the program is based - what are the rules and procedures by which we can get something as simplistic as a computer to create poems? It is worth noting that machine learning and data-oriented generative systems struggle to articulate these types of rules in a human-readable way.

previous / next