Welcome to the CHAI lab's page for election-related NLP!
This website features analyses of election-related texts. We compare the Democratic and Republican Party’s candidates’ agendas, candidates’ speeches, and national party platforms. We also analyze how the parties and candidates have changed their language since the previous presidential election cycle.
We currently employ two techniques: comparative summaries generated by large language models (GPT-4o) and log-odds statistics (Monroe et al., 2008) that reveal what words are more strongly associated with one corpus (e.g. the Democratic platform) relative to another (e.g. the Republican platform).
This website is an initial prototype for the 2024 U.S. General Election. We plan to expand this project’s scope to include state, local, and perhaps non-US elections, additional data sources (e.g. news and social media), and more sophisticated linguistic analyses.
We intend for this project to serve as a central resource for computational analysis of election-related content. We hope to make it as useful and informative as possible for computer scientists, political scientists, and non-academic audiences. We would love any feedback and suggestions!
Election 2024
Comparing the Democratic and Republican candidates and party platforms.
Then and Now
How have the candidates and parties changed since the last election?