Sentiment Analysis and Multiple Means Comparison for the 2020 United States Elections

Abstract: Here comes the abstract. Considering that the presidential elections between Trump and Biden have represented a great impact not only for the United States but also for the world and Mexico, in this work electoral preferences were analyzed using a natural language processing tool known as Sentiment Analysis. The methodology begins with reviewing and categorizing comments related to the 2020 US elections on the social network Twitter. Subsequently, the dictionaries are created to start with the sentiment analysis. In this way, three lines of analysis are established, being reflected in the following way: 1) data collection in the electoral campaign (information retrieval through downloads), 2) creation of dictionaries and 3) sentiment analysis. According to the previous order, first Tweets from different users have been randomly downloaded with the tagging algorithm, considering the comments of the Twitter attendees. The information seen as a corpus led to the definition of dictionaries and consequently, sentiment analysis bifurcates the information into two classes. Such categories have been called praise and name calling for convenience for the purposes of this article. Finally, the frequency of the terms is analyzed with descriptive and inferential statistics using the Fisher mean comparison.