Performing sentiment analysis to trace the history of identity and belonging in ancient Greek literature.

This article presents a Natural Language Processing (NLP) approach to analyze ancient Greek and Latin literature, focusing on sentiment analysis of words related to gender, race, religion, ethnicity, and other social identifiers. Using XML-encoded texts from large digital collections such as the Open Greek and Latin Project and the Perseus Digital Library, the authors developed Python libraries and trained machine learning models—including ancient Greek-BERT and Latin-BERT—to classify sentiments as positive, negative, or neutral in context. Preliminary results indicate that terms like "citizen" were generally used positively, while words for "foreigner" and "slave" carried more negative connotations in ancient Greek texts. The project aims to refine these models further and create tools for scholars to explore identity and bias in classical literature, demonstrating a collaborative potential between AI and the humanities.