Dissertation/Thesis Abstract

Detecting Genre in Greek Historiography Using Vocabulary and Logistic Regression Modeling
by Waller, Allyn, M.A., Tufts University, 2020, 47; 27957938
Abstract (Summary)

This thesis argues that there is a detectable, direct link between genre and vocabulary in Ancient Greek historiography. When split along canonical citation divisions and vectorized by a simple, non-lemmatized word count, a linear-regression model distinguished the historiographical works of Herodotus, Thucydides, and Xenophon from a counter-corpus of Demosthenes’, Lysias’, and Isocrates’ oratory, as well as Xenophon’s Socratic dialogs and Demosthenes’ letters. On 10-fold cross-validation, the model achieved overall accuracy of 95% ± 0.9% when these sections of text were labeled as either Historiography or Other, versus accuracy of 49.6% ± 2.7% when these sections of text were randomly assigned one of the two groups. This shows that vocabulary is a strong predictor for the genre of these texts. This thesis concludes with possibilities for application and expansion of these findings to other genres, time periods, and interpretation of the misclassified sections of text.

Supplemental Files

Some files may require a special program or browser plug-in. More Information

Indexing (document details)
Advisor: Crane, Gregory
Commitee: Hirsch, Steven, Proctor, David
School: Tufts University
Department: Classics
School Location: United States -- Massachusetts
Source: MAI 81/12(E), Masters Abstracts International
Source Type: DISSERTATION
Subjects: Classical Studies, Classical literature, Computer science
Keywords: Ancient Greek historiography, Genre, Historiography, Logistic regression, Modeling, Vocabulary
Publication Number: 27957938
ISBN: 9798645498191
Copyright © 2020 ProQuest LLC. All rights reserved. Terms and Conditions Privacy Policy Cookie Policy
ProQuest