Using machine learning to classify presidential candidate social media messages

Since presidential campaigns have incorporated social media into their strategic messaging, it has become more challenging for journalists to cover the election in depth, because of the large amount of data generated by candidates and the public every day. Journalists tend to focus on single quotes or tweets rather than providing analysis and reporting on the aggregate of messages on social media. But single tweets may not give people a full appreciation for the style of campaigning or the substance of the rest of the tweets. In order to get a sense of what the candidates and public are actually saying and how candidates communicate over time, we have taken a computational approach to predict categories of candidate-produced tweets and posts (as described in a blog post introducing the Illuminating 2016 project). We have been working on a system that automatically classifies each message into a category based on what the message is trying to do: urge people to act, change their opinions through persuasion, inform them about some activity or event, honoring or mourning people or holidays, or on Twitter having a conversation with members of the public.