Mrugank Dave, Bridget Donkor, Sam Elalouf, and Dennis Lee
We will be using Twitter’s API to gather data on tweets by politicians and the replies to / retweets of those tweets. We will then relate this data to various polling data regarding the politicians’ approval ratings.
Any politician faces the problem of how to represent themselves and their ideas or opinions to the public. As the last few years have made starkly clear, Twitter can be a powerful tool for politicians to communicate with voters. It can also be a useful way to gauge voters’ attitudes towards politicians. In this project, we will be measuring the affective response (and precursors) to certain tweets by politicians and relating those to shifts in their approval ratings over time.
We plan to analyze our data first by using sentiment analysis to characterize the affective response to each tweet. After building sentiment profiles for the reactions to the tweets we plan to use machine learning models to predict changes in approval rating based on changes in the affective responses to tweets.
Qualitative: We plan to use a number of visualizations in our project. We first plan to visualize the relative aggregate sentiments of responses to each tweet, then we hope to visualize the shifts in various polls estimations of politicians’ approval ratings.
Quantitative: After building profiles for the affective response to tweets and modelling the relationship between those response profiles and shifts in approval rating, we hope that we will see how accurate our model is at predicting shifts in approval rate based on affective response.