The Order of Things. A Study on Topic Modelling of Literary Texts.

Topic modelling is considered a statistical tool for the thematic decomposition of texts. In reality, it captures only statistical patterns in the structure of the object. This sensitivity of the method for structure makes it less effective when applied to literary texts in which structure itself is a relevant feature with an artistic function. In this paper, we calculate a series of topic models for three corpora of literary narratives with various stages of data cleaning. We apply coherence values, qualitative interpretation and measurement of topic distances in order to shed some light on the regularities between text features and the quality of the topic modelling performed for literary prose.