Experimental Creative Writing with the Vectorized Word - Allison Parish - TWIML Talk #72

Experimental creative writing leverages machine learning to transcend conventional poetic structures, moving beyond the mere imitation of human-authored verse. Allison Parrish, a poet and teacher at NYU, utilizes word vectors—mathematical representations of language—as a malleable medium for composition. By treating text as a continuous, multidimensional space rather than discrete units, she develops tools that allow for blending, blurring, and resampling language in ways that defy traditional narrative constraints. Her research extends into phonetic similarity vectors, which enable the analysis and generation of poetry based on sound patterns like slant rhymes and alliteration. This approach shifts the focus from achieving human-like verisimilitude to exploring the aesthetic potential of algorithmic procedures, emphasizing that understanding the underlying mechanisms of these tools is essential for interpreting the resulting artistic output.