Computational Stylistics in Poetry, Prose, and Drama pp 203–⁠222

Abstract

Procedures for generating poetry by recombining textual fragments give rise to very large spaces for solutions. While this is part of their charm, it also poses a significant challenge for readers trying to appraise their aesthetic merit. The present chapter proposes a computational solution to the exploration of such large spaces. The solution is based on sampling the search space in an informed manner driven by simple quantitative metrics about elementary features of poems. The quantitative metrics are selected empirically and validated in terms of their potential value for discriminating between different corpora of human-generated poems, poems generated by recombining human-written verse, and poems generated entirely by computational methods. This chapter postulates a mechanism for extracting valuable samples from the search space based on the establishment of a target vector of values for the identified features and applying a distance measure to the representations of this target and the features for each candidate poem in the corresponding vector space.

About

Gervás, P. (2023). N-Gram-Driven Word Level Recombination: Exploring a Search Space of Metrically Valid Verse. In A. S. Bories, P. Plecháč, & P. Ruiz Fabo (Eds.), Computational Stylistics in Poetry, Prose, and Drama: (pp. 203–222). Berlin, Boston: De Gruyter. doi: 10.1515/9783110781502-011

DOI
http://doi.org/10.1515/9783110781502-011

Print ISBN
978-3110-781-41-0

Online ISBN
978-3110-781-50-2

Published under Creative Commons Attribution 4.0 International License (CC BY 4.0)