A Comparison of Feature-Based and Neural Scansion of Poetry

TL;DR

This paper compares feature-based and neural methods for automatic poetic rhythm analysis, demonstrating that neural models with character-based representations outperform traditional features in English and Spanish, emphasizing the importance of whole word structure.

Contribution

It introduces a neural approach using Bi-LSTM+CRF for poetry scansion, showing superior accuracy and highlighting the value of whole word information over syllable-level features.

Findings

Neural models outperform feature-based methods in scansion accuracy.

Character-based representations are more informative than hand-crafted features.

Whole word structure significantly improves poetic rhythm analysis.

Abstract

Automatic analysis of poetic rhythm is a challenging task that involves linguistics, literature, and computer science. When the language to be analyzed is known, rule-based systems or data-driven methods can be used. In this paper, we analyze poetic rhythm in English and Spanish. We show that the representations of data learned from character-based neural models are more informative than the ones from hand-crafted features, and that a Bi-LSTM+CRF-model produces state-of-the art accuracy on scansion of poetry in two languages. Results also show that the information about whole word structure, and not just independent syllables, is highly informative for performing scansion.