Abstract We explore the classification of ancient Greek literature using computational semantic methods, in order to compare traditional genre categories with data-driven clustering. We rely on the Diorisis corpus and apply word embeddings and then hierarchical clustering to study semantic similarity between works. While ancient and modern classifications are grounded on formal, thematic, or ethical criteria, our results reveal partial but imperfect alignment: poetry and theatre cluster clearly, whereas prose subgenres like oratory, philosophy, and technical treatises intermingle. After careful examination, these outliers in the semantic clustering align with the more specialized scholarship on these works. We argue that computational modeling can both illuminate and critique inherited classifications, offering classicists an empirical basis for organizing ancient texts.
Laurent Gauthier (Wed,) studied this question.