Gender Differences in the Rhetoric Used in Political Advertisements
|
My dissertation examines political rhetoric on criminal justice adopted by legislative candidates in their political ads. In the first empirical chapter, I focus on the impact of gender, party identify, and their interaction on candidates' political rhetoric on criminal justice policy. I utilize Google’s Speech-to-Text API to transcribe more than 1100 political ad videos in 2012 and 2016 into text data and then conduct automated content analysis. Using structural topic models, I find the topic of sexual assault and women’s rights were more likely to appear in the political ads sponsored by women than those by men. While Democrats frequently talked about sexual assault and women’s rights, Republicans seldom touched this topic. In addition, women were even more passionate about this topic than men within the Democratic party but we do not see such a difference within the Republican party. Most interestingly, women candidates are less likely to highlight deviants such as illegal immigrants and drug dealers in their political ads. I also find there was a marked difference in the way deviants were discussed between men and women – negative words such as “radical” and “terror” were more associated with men than with women.
|
|
Automated Analysis of Campaign Advertising on Facebook
With Frederick Corpuz, Lance Lepelstat & Erika Franklin Fowler
|
In this study, we apply both automated image and textual content analysis to a novel, non-traditional data set - Facebook political ads of twenty-four senatorial candidates in the 2018 midterm elections. To conduct automated analysis of such data, we use machines that can both "see'' and "read.'' An online political ad usually consists of three parts: body text, ad image and image text. The figure on the left shows two examples of our data. Using state-of-the-art deep learning algorithms, we recognize the appearance of top leaders such as Trump and McConnell in the panel (a) as well as senate candidates and their opponents such as Rosendale and Tester in the panel (b).
|
|
Facial recognition has been one of the most researched topics in artificial intelligence. As traditional machine learning methods have been superseded by deep learning methods based on convolutional neural networks (CNNs) since 2012, face recognition has indeed become one of the most successful applications of computer vision. With few exceptions, however, it is rarely employed in political science. In this study, we use a deep-learning-based facial recognition tool to determine the appearances of political leaders, candidates, and opponents in Facebook political ads. This approach is especially suited for online political ad images because faces in them are purposefully well presented, while face images in the real world, or faces-in-wild, are more likely to be of poor quality due to low resolution, dim light condition and occlusion.
|
|