Sentiment Analysis of Tweets Including Emoji Data

Sentiment analysis of text is a valuable tool used to identify and classify bodies of text for various purposes, including public sentiment detection in political campaigns, spam detection and threat assessment. In this paper, we examine the effectiveness of incorporating Emoji data for Twitter data emotion classification. We conducted experiments using Multinomial Naïve Bayes (MNB) and Support Vector Machine (SVM) classification methods, with automatically labeled Twitter data. We compare the accuracy of these classification methods with and without the Emoji data included over varying vocabulary sizes. We find that MNB outperforms SVM on the data for large vocabulary sizes, and both classifiers perform slightly better with the Emoji data included.