Authors:

W. Bowman

Date Created:

Not specified

Course Title:

Professor:

Not specified

About Paper:

Recent Artificial Intelligence developments initiated discussions on whether and how we could apply technologies to classical studies. Though great strides have been made in the area of natural language processing in ancient texts, specifically with the aspect-extraction of Greek-Bert, most researchers have found it difficult to apply their software to text that lies outside of their sample data. Further, there has been little progress on a sentiment analysis model for Ancient Greek and Latin texts. Our research aims to bridge this gap. Specifically, we use data from open source libraries such as the Open Greek and Latin Project and the Perseus Project, allowing us to compile the majority of surviving Ancient Greco-Roman texts. For software development, we employ Greek-Bert for aspect tagging, Classical Language Toolkit for lemmatizing, and Python Aspect-Based Sentiment Analysis Package for sentiment labelling. Additionally, we construct testing data using random sentence generation by providing lists of words and connecting them to form rudimentary sentences with associated sentiments. At this point, we have developed an NLP model that is beginning to understand Ancient Greek and Latin beyond their linguistic structure, and, even though our results are preliminary, the model is broadly able to differentiate negative and positive sentiment associated with a sentence aspect. This work allows us to quantify the way in which ancient peoples discussed different races, genders, and ethnicities.

Source:

University of Florida / W. Bowman, D. Chong, G. Gnanam, M. Gontu, Z. / 2023

Topics:

No topics listed

Co-authors:

W. Bowman