My Ever Changing Moods: Sentiment-Based Event Detection on the Cloud

Abstract

Twitter is a globally used micro-blogging platformwith hundreds of millions of tweets sent every day. Manyresearchers have explored Twitter analytics across a wide rangeof areas such as topic modeling, sentiment analysis, eventdetection, as well as the application of Twitter for a variety ofdomain-specific application areas, e.g. disaster management. Onearea that has not been explored is how changes in sentiment canbe used to identify events. In this paper we present a scalableCloud-based platform for harvesting, processing, analyzing andvisualizing large-scale Twitter data. We focus especially on howchanges in sentiment can be used to identify events in givencontexts. What is novel is that the events that are detected are notdependent explicitly on the topic of any given tweet, but entirelyon the change in sentiment. This offers new capabilities for eventdetection that have hitherto not been explored. To illustrate theapproach, we present case studies related to sporting eventsidentified entirely through changing sentiment with specific focuson the 2014 FIFA World Cup of Soccer and the 2015 World Cupof Cricket.