Topic Modeling and Classical Latin: a Large-Scale Corpus Investigation

The objective of this thesis is to conduct a comprehensive study that applies topic modeling to a corpus of Latin literature and to assess how topic models, through their capacity to predict latent thematic structures and patterns, can enrich the understanding of this vast and diverse literary tradition. The results demonstrate that the neural topic models, trained on the datasets considered, effectively predict coherent and interpretable sets of topics, showcasing their potential for examining the thematic structures of ancient texts. Indeed, the topics obtained are thoroughly analyzed from a literary standpoint, offering an innovative way to interpret key developments in the history of Latin literature and revealing new dimensions in the works of the authors included in the corpus.