Sentiment Analysis on Tweets with Punctuations, Emoticons, and Negations

Abstract

Social media allows people to instantly communicate and share information with each other in a public forum. Twitter is a social media website where people compose short text messages (commonly known as tweets) which contains emotions and feelings. Furthermore, Twitter is considered to be a large repository of emotions, sentiments, and moods. Prevailing studies show that analysis of sentiments can be done through machine learning. Algorithms, features, and tools may be of use to train a model to classify sentiments of a tweet. In this study, tweets which includes different parameters (emoticons, negations, and punctuations) were gathered and annotated by experts to label their sentiments. Machine learning was applied in this study in order to formulate an optimal model. The results of the experiment shows that the features included provided a significant performance in order to identify the sentiment of a given microblog statement. The algorithms used to build the model are KNN and Naive Bayes with on both English and Filipino datasets.