Polarity Classification on Twitter- You are “What you Tweet”

Information-gathering has always been an important part to find what other person is thinking. Millions of users tweet on different aspects of life every day. Therefore micro blogging websites are a very good source for polarity classification. We introduce a novel approach which automatically classifies the polarity of Twitter message. These messages so called tweets are classified as positive or negative or neutral. These results are useful for the customers or any general user who wants to research about the polarity of products before purchases, or it can be useful for the companies that want to analyze the reviews from people of their brands in the market. Most of the previous research on classifying the polarity of messages has tried to achieve some good results but have ignored the neutral tweets which lead to wrong polarity classification, so we have tried to solve this issue in our project. We present an approach for classifying the polarity of tweets using machine learning algorithms using a novel feature vector. Our training data contains publically available tweets which are obtained using twitter API’s available. The following report shows the steps for preprocessing the dataset to achieve high accuracy. The novel feature vector of weighted unigrams that are used to train the machine learning classifiers is the main contribution of our project.