Survey of sentiment analysis of political content on twitter

Abstract

Social media usage has seen a dramatic rise in the recent years. With the use of just 140 characters on Twitter, people can voice their opinion on any subject. Various techniques have come up to identify the sentiment of these tweets so as to reach appropriate conclusions. Sentiment analysis of tweets related to politics is theorized to be able to identify public sentiment toward candidates and predict election results. The methodologies have been broadly classified into lexicon-based approaches and machine learning-based approaches. Algorithms like Naive Bayes (NB), support vector machine (SVM) and neural networks are used in machine learning-based approaches. Owing to their greater flexibility, they are more useful than lexicon-based approaches when it comes to political tweet analysis. A survey of all these approaches reveals that SVM provides accuracy over 70%, making it the most efficient algorithm for political sentiment analysis of tweets.