Data Mining in Presidential Debates and Speeches: How Campaign Rhetoric Shaped Voter Opinion in the 2016 U.S. Presidential Race

Abstract

The goal of this thesis is to understand how President Donald Trump’s rhetoric impacted the opinions of various voter groups throughout his campaign for the 2016 U.S. Presidential Election. We use natural language processing techniques to investigate the linguistic features of his rhetoric in the three presidential debates and the sixty-four speeches he delivered after winning the Republican presidential nomination. In our analysis, we compare the linguistic features of Trump’s words to that of Clinton’s words and find that Trump’s words were typically more common in an American English corpus and more extreme on both ends of the sentiment spectrum. We also find that Trump not only used rhetorical devices for persuasion but also adeptly coupled these devices with the right talking points based on the composition of his audience. Precise execution of this strategy garnered him an unexpectedly large number of votes from the white female and Hispanic demographics.