A Quantitative Performance Evaluation Of Machine Learning Algorithms For Analysing Sentiments Of Emoticons

Twitter is one among the most popular platforms for people to express their opinions on a certain issue or product. As it consists of a multitude of users, Twitter is a significant source of data for companies to analyze product reviews, or general public opinion. On Twitter, a tweet can consist of 280 characters. Due to this limitation, users have resorted to using emoticons in a more efficient fashion. This has been achieved by replacing multiple words that indicate an emotion, with a single emoji. Traditional sentiment analysis algorithms and classifiers are highly efficient for usage on text data that does not contain any emojis. This paper evaluates the existing sentiment analysis algorithms in their capability to analyse emoticons in text.