On Segment Size in Poetry Analysis Using the Latent Dirichlet Allocation Method

Abstract

The purpose of this study is to examine one of the problems in applying latent Dirichlet allocation (LDA), to poetry works, and to examine the effectiveness of LDA in poetry research. Not only LDA, but also topic modeling in general is considered a promising approach in the field of digital humanities and text mining, and although the number of studies using topic models to study prose works has been increasing in recent years, there have been few studies applying topic models to poetry works. This paper uses the poems of Alfred Tennyson, a Victorian poet, as a target corpus to point out the challenges of applying LDA to verse texts, and to discuss the feasibility of running LDA on texts split into smaller chunks of an equal size.