Applications of twitter emotion detection for stock market prediction

Abstract

Currently, most applications of sentiment analysis focus on detecting sentiment polarity, which is whether a piece of text can be classified as positive or negative. However, it can sometimes be important to be able to distinguish between distinct emotions as opposed to just the polarity. In this thesis, we use a supervised learning approach to develop an emotion classifier for the six Ekman emotions: joy, fear, sadness, disgust, surprise, and anger. Then we apply our emotion classifier to tweets from the 2016 presidential election and financial tweets labeled with Twitter cashtags and evaluate the effectiveness of using finer-grained emotion categorization to predict future stock market performance.