Classification for Catharsis: Predicting the Authorship of Ancient Greek Tragedy

Abstract

The issue of dubious authorship, persistent for centuries in discussions of classical literature, has been enhanced in recent years by the use of machine learning techniques. Statistical classifiers such as naive Bayes, support vector machines, and logistic regression have shown remarkable accuracy in ascribing documents of varying lengths to the right authors and have thus been central in parsing the extent of potential interpolations in ancient Greek literature. This thesis seeks to expand on this recent trend by designing and applying a new algorithm based on generalized linear mixed models, evaluating its performance against standard authorship attribution methodology. In particular, these models will be run on select works by the dramatist Euripides that have not yet been analyzed through this statistical lens.