Social media represents a new mechanism by which individuals form opinions about social and political events. Most social media users now have the ability to communicate with other users, even if they are separated in geography and ideology. As an initial step in studying communications on social media, we present a workflow in R to detect events. Specifically, we use latent Dirichlet allocation (LDA) of tweets from Twitter at well-defined time intervals to detect important political and social events. A critical step in our workflow is the data wrangling of raw tweet downloads from Twitter, which we accomplish with the new parseTweetFiles R package. We illustrate our methods on tweets near and during the time of the National Football League’s 2015 Super Bowl game. With this collection of tweets, we detect short-lived topics related to 1) important Super Bowl plays and players and 2) Super Bowl half-time show performers. We then discuss implications of our methods for event detection and place our findings in the context of scholarly discussions of social media discourse.
