Date of Award

5-2018

Degree Type

Thesis

Degree Name

Master of Science (MS)

Department

Computer and Information Technology

Committee Chair

Julia Rayz

Committee Member 1

J. Eric Dietz

Committee Member 2

John Springer

Abstract

This research aims at exploring the polarization in the news based on reports of the 2017 “Travel Ban” executive order using natural language processing and clustering techniques. The study uses a 2014 report from Pew Research Center as a source of perceived relative positioning of various mainstream news sources on an ideological scale. This positioning is compared to the relative positioning and grouping of news articles from selected news sources; the positioning is revealed by the analysis of their news content. The dataset for this study comprises of 2178 news articles about the 2017 executive order on immigration, commonly known as the “Travel ban,” published during 27th Jan. to 15th Oct. 2017 in the selected sources. K-means clustering results for word n-grams and paragraph vectors are compared. The better performing paragraph vectors approach is then used for hierarchical agglomerative clustering using Ward’s linkage method. The clustering quality is evaluated using purity and normalized mutual information metrics. Results show that the articles from the same author are closer to each other than the articles from different authors. Among the articles from different authors, the articles from same news source are closer to each other than those from other news sources. The relative positioning of news sources based on their mean paragraph vectors is consistent with the relative distance among different news sources outlined in Pew Research Report on positioning of mainstream news sources on the ideological scale.

Download

Share

COinS