Automated Pipeline for Sentiment Analysis of Political Tweets

Abstract

Social media such as Twitter have proven to be great resources of information regarding many events that proliferate the entire world. It also has the power to change the opinions of millions which is especially useful to sway the masses during political campaigns like presidential elections. Sentiment analysis can be performed on tweets to determine how people feel about certain political events which can be used to predict the behavior of people and propagate real-time change as the events play out. Therefore we decided to create a sentiment analysis and polarity detection pipeline generalized for all political data. The pipeline attempts to automate all the NLP tasks from data scraping to cleaning and pre-processing the dataset to make it ready for the classification tasks. The predictions are visualized via word clouds and a map color coded to reveal the sentiments of key nations around the world regarding the political event. This pipeline is tested from end-to-end with our own personal use case being to determine which candidate do countries around the world preferred during the 2020 US Presidential Elections: Trump or Biden. The pipeline provides fruitful results with a test accuracy of 73.73 percent.