Emotion detection from “the SMS of the internet”

Due to the sudden eruption of activity in the social networking domain, analysts, social media as well as general public are drawn to Sentiment Analysis domain to gain invaluable information. In this paper, we go beyond basic sentiment classification (positive, negative and neutral) and target deeper emotion classification of Twitter data. We have focused on emotion identification into Ekman's six basic emotions i.e. JOY, SURPRISE, ANGER, DISGUST, FEAR and SADNESS. We have employed two diverse machine learning algorithms with three varied datasets and analyzed their outcomes. We show how equal distribution of emotions in training tweets results in better learning accuracies and hence better performance in the classification task.