Predicting metrical patterns in Spanish poetry with language models

TL;DR

This study evaluates the effectiveness of BERT-based language models in identifying metrical patterns in Spanish poetry, comparing them to existing automated systems through extensive experiments.

Contribution

It demonstrates that BERT models, originally designed for semantic tasks, can effectively capture structural poetic features in Spanish, offering a new approach to poetic metrical analysis.

Findings

BERT models perform comparably to specialized metrical systems.

Fine-tuning improves the accuracy of language models for poetic pattern recognition.

Structural information is retained in BERT models despite their semantic training focus.

Abstract

In this paper, we compare automated metrical pattern identification systems available for Spanish against extensive experiments done by fine-tuning language models trained on the same task. Despite being initially conceived as a model suitable for semantic tasks, our results suggest that BERT-based models retain enough structural information to perform reasonably well for Spanish scansion.