Sentiment analysis of tweets on social security and medicare

Abstract

The field of politics can be greatly aided through the use of natural language processing techniques on places like social media. Twitter hosts extensive discussions on political topics like Social Security and Medicare. We gathered almost 90,000 tweets on Social Security and Medicare for sentiment analysis. Positive sentiment was higher than negative sentiment, and public discussion was polarized with relatively little neutral sentiment. We conducted named entity recognition and found that entities of people and organizations were most prevalent in discussion regardless of sentiment. Topic modeling was also performed, and we determined the dominant topics in tweets. We compared word counts of keywords to the probabilities that each belongs to their respective topic in order to gauge the impact of keywords in their topics. The discussion of Social Security and Medicare on Twitter is objectified through named entity recognition and topic modeling. The experimentation conducted can be broadly applied to politics to better understand objects and themes of key interest in various complex issues that are debated and discussed on Twitter. This study provides a comprehensive structure to the public discussion of Social Security and Medicare on Twitter and assists politicians and lawmakers in making significant, relevant decisions and policies.