PoetaQuae

Latin text classification

By Mackenzie Mitchell and Brendan Gilroy

Overview

The goal of this project is to create a machine learning program to categorize what poet/author wrote a given classic Latin text.

Data

Data was web scraped using BeautifulSoup from:

• http://rudy.negenborn.net/catullus/catullus2.htm
• https://www.sacred-texts.com/cla/virgil/
• http://www.perseus.tufts.edu

As observed, Cicero has many more data points (or documents) than the other poets. This would highly bias our classifier so we decided to remove Cicero to create more balanced classes:

Once Cicero was removed from the dataset, the classes were much more balanced:

The dataset contains:

• 118 personal/lyrical poems by Catullus (84-54 BC)
• 3 epic poems by Virgl (70-19 BC)
• 16 satirical poems by Juvenal (Late 1st Century-Early 2nd Century AD)
• 2 lyrical peoms by Horace (65-8 BC)
• 1 mythological peom by Ovis (65-8 BC)
• 1 historical prose piece by Caesar (100-44 BC)

Looking at the most frequenty used stop words in all of our corpus:

• -que : and
• -ne: negates statement and turns sentence into a question
• ex: Est Catullus: Catullus is here.
• ex: Estne Catullusque Cicero: Are Catullus and Cicero not here?