Changing the Subject: Dynamic Discussion Monitoring in Twitter

Abstract

In recent years, Twitter has become increasingly popular both as a social networking service (where users express their opinions) and as a tool for information retrieval. Many events that are commented, debated or argued online, however, are dynamic and unpredictable in nature, resulting in the need to derive the corresponding dynamic computational methodologies to track and to extract such changing topics, events and relevant content in a timely and unattended manner. In this paper, we propose a framework to accomplish two objectives: periodically obtaining the relevant topics of discussion among authorities and adapting the tracked keywords or hashtags accordingly to retrieve the most relevant possible information in each time window. The application of this framework to a case study reveals how our proposed approach is able to accurately and dynamically track the conversation around an event. Our method could therefore be applied to query Twitter in many domains, such as politics, sport events, marketing campaigns or media engagement, among others.