Predicting and Classifying the Poetic Quality of the Early and the Late Generation of Romantics: A Computational Stylistic Study

Abstract

Poetic quality is an important value for a poet’s literary craft, and it is considered as a cultural heritage of a society. Poetry is made with language and poets use language features in a particular way to create effects and to convey meaning. Measuring poetic quality would appear to be an intrinsically qualitative endeavor. However, it is possible to objectify some of the language features to some degree using a quantitative criterion, which is here proposed. The study represents a preliminary literary computational approach in relation to Romantic poetry which is dominated by a few names: William Blake, William Wordsworth, Samuel Taylor Coleridge, Percy Bysshe Shelley, John Keats, and Lord Byron. There has been little or no research with literary computing techniques focused on measuring and classifying the quality levels of those poets’ works. The study builds on a set of ten measurable stylistic features extracted from one hundred and ten poems regardless of genre or subject. The study carries out an analysis using Binary Logistic Regression Modeling (BLRM) as one of the most commonly used predictive modeling techniques. I used the stylistic features as predictive variables and the quality as the outcome variable indicative of the stylistic characteristics of each of the poems selected. In this way, this study is able to measure the stylistic features in the selected poems and, independently, using Binary Logistic Regression to predict the general poetic quality levels for each of the poets examined. The study is also able to make connections between some of the poets which enabled a more detailed view of the subcomponents’ usage and occurrence in the poems.