Computational Approaches in the Humanities: from Sentiment Analysis to Deep Learning Colorization /

Bibliographic Details
Main Author: Kim, Hoyeol (Author)
Other Authors: Ives, Maura (Thesis advisor)
Format: Thesis eBook
Language:English
Published: [College Station, Texas] : [Texas A&M University], [2023]
Subjects:
Online Access:Link to OAKTrust copy
Description
Abstract:This dissertation is article-based, consisting of four chapters with two themes: colorization (chapters 1 & 2) and sentiment analysis (chapters 3 & 4). Chapter 1, ⁰́₋Victorian400: Colorizing Victorian Illustrations,⁰́₊ reveals my methods for creating, curating, and validating the Victorian400 dataset for colorizing Victorian black-and-white illustrations. Victorian400 is a nineteenth century illustration dataset consisting of 400 colorful images that is helpful for testing and developing deep learning models. I tested the Victorian400 dataset with the pix2pix model, which is a conditional generative adversarial network (cGAN) model, to verify the dataset for colorizing black-and-white illustrations from the nineteenth century. Chapter 2, ⁰́₋Case Study: Using Machine-Colored Illustrations of Charles Dickens⁰́₉s Fiction in the Classroom,⁰́₊ addresses the pedagogical usage of machine-colored illustrations in the English classroom based on my case study. I discovered that students often preferred either no illustrations or machine-colored illustrations to hand colored ones. In chapter 3, ⁰́₋Sentiment Analysis: Limits and Progress of the Syuzhet Package and Its Lexicons,⁰́₊ I explore the limits and progress of the Syuzhet package, a sentiment analysis tool in R, for the sentiment analysis of literary texts. I compare the four sentiment lexicons (Syuzhet, Bing, Afinn, and NRC) used in the package. I also test Syuzhet, SentimentAnalysis, sentimentr, RSentiment, and VADER (Valence Aware Dictionary and sEntiment Reasoner) with seven different sentences to see how each lexicon-based sentiment analysis tool generates sentiment scores. In chapter 4, ⁰́₋Dickensian Sentiment and Sentiment Analysis of Victorian Novels,⁰́₊ I perform sentiment analysis on three Victorian novels using BERT (Bidirectional Encoder Representations from Transformers) in tandem with a dataset I created for the sentiment analysis of Victorian fiction, in order to see if sentimentality is revealed in Charles Dickens⁰́₉s Our Mutual Friend, George Eliot⁰́₉s Middlemarch, and Charlotte Bront©±⁰́₉s Jane Eyre, since sentimentalism was a primary social value in Victorian society. The electronic version of this dissertation is accessible from https://hdl.handle.net/1969.1/197350
Item Description:"Major Subject: English"
Includes vita.
Physical Description:1 online resource.
Bibliography:Includes bibliographical references.