The Usage of Twitter Data for Early Crisis Detection

Abstract

The establishment of novel information and communications  technologies, such as social media, has changed how individuals and  organizations behave. Particularly during uncertain situations caused by natural  or man-made disasters, people communicate and share information via social  media. Emergency services could use published content of social media  participants for an early detection of crisis situations and thus enhance crisis  management. However, exploiting these information bears several challenges.  To gain insights from large amounts of unstructured data in time-critical  situations, emergency services are in the need of tools that enable them to  gather, process, and analyze relevant information. In this research article, we  follow the design science paradigm and propose a concept (artifact) of an early  crisis detection tool. Based on expert interviews, design implications for the  concept tool are derived. In addition, we conducted a design thinking workshop  to evaluate the artifact, highlighting its practical relevance and usability.

Introduction

The continuous development of new technologies in recent years has changed our  everyday lives profoundly. Mobile technologies and embedded computing devices, as  well as network and communication services, such as Facebook and Twitter, have  changed the way how people communicate and share information with others [1].  Social media has become an important medium in particular during disaster situations  [2, 3]. There is a growing consensus about the fact that social media can be valuable for emergency services in all disaster phases [4–7]. But other than the use of social media in crisis response, only little is known about the use of social media for  Emergency Management Agencies (EMAs) in the early detection of a crisis [4–6]. By having the knowledge of how to detect a crisis and how to gather these data in time-critical events, unleash the potential to minimize the damages and possibly save lives.  Furthermore, regarding the massive amount of unstructured data available in social media, it becomes apparent, that EMAs need technological tools to support crisis  management efforts by automatically collecting, aggregating and visualizing  important information [8]. Regarding existing tools intended to solve these  challenges, [9] reported that the state of the art in the field of social media-based crisis  management systems comprises a multitude of heterogeneous, domain-specific  solutions instead of general applicable, modular and flexible tools. Thus, the design of  a concept for a comprehensive crisis detection tool, deployable in the control center of  EMAs, is one major research aim of this work.

To reach the mentioned research aim of developing an innovative IT artifact in  form of a concept for an early crisis detection tool, we seek to address the following  general question. How can Twitter data be used for an early crisis detection tool?   

First, we investigated which Twitter data can serve as indicators to early detect a  crisis. Second, we studied the design of an early crisis detection tool, by investigating which features the tool should comprise for processing, analyzing and visualizing the  gathered data. Hence, this work starts with the research design following the design  science approach [10]. The design science research methodology (DRSM) by [11] is  further used for structuring this paper. Accordingly, an unsolved problem will be identified and the motivation to close the defined research gap will be presented.  Afterwards, the objectives of the newly designed IT artifact indented to solve the  identified problems, will be defined. To assess the state-of-the-art of current literature  as well as practical demands, the chosen research methods are a literature research  and expert interviews. Based on the results, the design and development of the tool  will be described. In a further step, the tool will be demonstrated and evaluated by  experts in the field of crisis management during a design thinking workshop and in a second step during a test scenario including three participants testing the tool in a  two-day period in the emergency management context. Last, we discuss the results  and point out relevant aspects for further research.