Sentiment Classification of Tweets with Non-Language Features
J. Akilandeswari, Jothi Ganesan · Procedia Computer Science · 2018
In recent years, mining social media sites like Twitter, Facebook have become a hot research focus. Twitter is one of the most fashionable microblogging services that permit users to express their views on e-commerce websites, sports, politics, modern technologies, movies, spirituality and so on. Sentiment can be considered as a natural expression of viewer’s perception. It is extremely difficult to identify the sentiment/opinion about a specific product or event by collecting and compiling microblog tasks manually. It is only fair to develop a system which collects, compile and analyse microblog tasks to arrive on insights that helps to take an action against an event. The system can monitor and evaluate in real time online views, to demonstrate how the whole world is reacting to a concept/ideology/event. Developing such a system which assigns polarity to a tweet is a hard task. In this paper we propose a scoring methodology to find the sentiment polarity of the Twitter messages. Emotions, shortened words and non-language features are integrated to increase the significance of the score computed for assigning the polarity for the tweets. The experimental results show that the proposed method enhances the accuracy of the assignment of polarity to tweets.