Presidential Debates

I think we all know that the first 2024 US presidential election debate was a little wacky. This got me thinking about past presidential debates, which apparently have an interesting history. I wanted to measure the topics brought up by each candidate and compare them to the topics the audience wanted to hear about, but it mainly turned into a generic NLP analysis of all of the presidential debates since 1960.

I got the transcripts for debates 1960-2020 from https://www.kaggle.com/datasets/bahdahshin/presidential-debates, it included video files for each debate as well which I didn't need, so I re-released the dataset here with only the text transcripts. I didn't change anything else except some file structure.

This is a living document, I'll probably add more later, but for now it's a basic sentiment analysis of all debates 1960-2020 (I'll add the 2024 debates when all of them are finished).

Sentiment Analysis

I used nltk's inbuilt SentimentIntensityAnalyzer to gauge the sentiment of the debates (after cleaning the text). Here's the sentiment analysis chart (meausred by negative, neutral, and positive sentiment).

Sometime in the future I will probably do more analysis, and analyze why they had the characteristics that they did.