Articulations and Compasses

Most methods for computationally calculating similarity among texts rely on statistical models of semantics, but meaning has always seemed like a secondary feature of poetry to me. (My lowest scores on standardized tests as a kid were always in the "reading comprehension" category.) The techniques I used to write Articulations resulted from experiments in designing methods to calculate text similarity that focus instead on phonetics and syntax. The book in its final form has two sections, both computational cut-ups of lines of poetry scraped in bulk from Project Gutenberg: "Tongues," a long prose poem where lines are arranged in a way that maximizes phonetic similarity from one line to the next, and "Trees," a series of twenty-five poems composed of lines that share syntactic characteristics. Compasses, a chapbook recently published in Andreas Bülhoff's sync series, consists of poems produced with the help of a machine learning model I designed as the next step in my exploration of phonetic similarity. In addition to being able to represent the phonetics of a text and find phonetic similarities among texts, this model can generate entirely new texts based on arbitrary phonetic representations. In Compasses, I used this model to generate new imaginary words that exist in the negative phonetic spaces between the names of members of well-known quartets.