Threat Emotion Analysis in Social Media

Abstract

In recent decades, the internet access growth has generated a substantial increase in the information circulation in social media. Within the information variety circulating on the internet, extreme social events such as armed conflicts have become areas of great public interest because of their direct influence on society. The study of such data from social media is useful in understanding an event's evolution, in particular how threats over time can generate an endogenous evolution resulting in an extreme event. This chapter uses the technique of sentiment analysis to identify the threat degree of news about armed conflicts distributed in social media. This analysis generates an endogenous threat time series that is used to predict the future threat variation of the analyzed extreme social events. In the prediction of the endogenous time series, the authors apply the deep learning technique in a structure that uses the long short-term memory (LSTM) neural network.