TweetFeel: Analyzing Emotions in the Twittersphere

The majority of people on the planet now have access to the Internet for text, image, audio, and video communication. Through social media, people from many backgrounds share knowledge about current events and express their own opinions on them. Analyzing people’s emotions is necessary to comprehend and identify the behavior of such vast amounts of textual data about them. The study focuses on information gathered from Twitter, one of the most widely used social media platforms, by examining both historical and real-time feeds and extracting emotions from them. The necessary English-language Twitter data is transformed into a vector of six emotions, and supervised learning methods like Logistic Regression, Naive Bayes, SVM, Decision Tree, Random Forest, Gradient Boosting, and Multi-Layer Perceptron are utilized to identify the label of one of the basic human emotions. Gradient Boosting and SVM are the top-performing models that reported the highest F1-Score of 0.88 among the other models. Sadness and joy came out to be the best-performing emotions whereas it was comparatively difficult to predict surprise emotion.