TWITTER SENTIMENT ANALYSIS USING UNSUPERVISED AND SUPERVISED APPROACH : A CASE STUDY

Abstract

Opinion analysis or sentiment analysis is one of the most sorted out technique these days in order to determine the sentiments or emotions of people regarding any event. This technique has come into existence because of extensive use of social media platforms like Facebook, Twitter etc by people to express their emotions regarding any event that has occurred or any event that is most likely to happen, be that the release of a movie or a political rally that is about to take place. People make sure to express their sentiments. Sentiment analysis proves very beneficial for any company selling a product to know how their product was received by people or by any political party to determine how people are reacting towards their running candidate. Different events occur worldwide so it is not very easy to determine the emotions and sentiments of people regarding these events; it results in huge amount of data and many steps to reach any conclusion about the sentiment. In order to analyze these sentiments two approaches of machine learning can be used – unsupervised or supervised. Machine learning algorithms can be used to determine whether a series of words reflect a positive, negative or neutral meaning. Unsupervised learning involves a rule based or lexical approach and this can be done using the pre-built open source libraries like TextBlob, VADER. Unsupervised learning is a simple and efficient approach that has been used in the past so many years to determine the sentiments or opinions, over the years many libraries have been built to ease the task of analyzing the sentiments. This library is used to determine whether the sentiments are positive, negative or neutral effectively. Once the sentiments have been recorded and the data is converted to a more structured form, these can be fed to the machine learning algorithms like Naïve Bayes, SVM, LSTM and so on in order to predict the accuracy with which these algorithms can predict the sentiment.