Divining Language: Unearthing Medieval Italian through NLP

Abstract

The use of Natural Language Processing (NLP) allows for analyses on texts that extend beyond mere semantic interpretations. However, the application of these methods to historical texts, particularly medieval Italian, poses significant challenges because of the linguistic evolution that results in widespread differences between modern and medieval Italian. This thesis addresses this gap in NLP applications for medieval Italian by developing methodologies to process and analyze texts from this period. Our research focuses on Dante Alighieri’s Divina Commedia, looking to enhance the understanding of this work through NLP techniques. We adapt existing NLP tools to accommodate the linguistic features of medieval Italian, enabling more accurate text processing and analysis. Using a combination of co-occurrence networks, topic modeling, and vector embeddings, we aim to reveal underlying thematic and semantic structures within the Divina Commedia that might have been less visible when using traditional analysis methods. This work not only serves to create a toolset that can be used on Dante’s work, but also establishes a framework for future NLP applications to other medieval Italian texts.