SUPER is a joint effort of social media and security experts towards introducing a holistic, integrated and privacy-friendly approach to the use of social media in emergencies and security incidents
SUPER explores a holistic integrated framework for understanding citizens’ reactions against emergencies in social media, while at the same time empowering security forces and civil protection agencies to fully leverage social media in their operations
Natural and human-made disasters are traumatic events which leave people feeling terrified, powerless, and/or horrified in the face of threatened or actual injury or death, overwhelming their responsiveness. Hence, security and/or emergency management has to take into account these psychological reactions and integrate them. Furthermore, the understanding of crowd behavior is crucial, in order to prevent and manage emergency situations. SUPER will part from leading behavioral theories on citizens’ individual and collective reactions to emergency, disaster and crowd situations. It will mainly concentrate on psychological responses during and after traumatic events and physical crowd behavior. Despite the vast literature on Acute Stress Response, Post Traumatic Stress Disorder, and crowd behaviour, there only exist few and partial studies of social media uses during and after traumatic events. SUPER aims to fill in this gap and model citizens’ social media-behaviour in relation to emergencies.
One of the key challenges to leveraging social media in emergencies is tackling information veracity, since much of the information posted within social media can be incomplete, incorrect or malicious. Rumour tracking aims to provide information verification technologies that will enable response services to safely use social media to better understand an event and make decisions. SUPER will leverage state-of-the-art supervised machine learning technologies to determine the veracity of rumors within social media and to surface reliable information to emergency responders.
To make effective use of social media, technologies that facilitate fast and targeted information gathering are needed. SUPER will develop real-time search tools that enable emergency response teams to rapidly explore the information provided by different high volume social media sources. The real-time search component of SUPER will build upon the Terrier information retrieval platform (http://terrier.org) with efficient distributed search over parallel social media streams, facilitating low latency access to key information during incidents and emergencies.
Topic Communities play a vital role in understanding and dealing with emergency management. Our goal is to detect, categorize and classify the people who have common needs, behaviors and acts based on their actions and texts in order to help them immediately in cases of emergency. SUPER will use state of the art techniques to discern user topic communities and track their evolution quickly and with high precision. This will allow different types of people to be helped during and after the emergency situation according to their needs.
The growing popularity of virtual spaces (such as social networks) among heterogeneous groups of the population makes them an ideal testbed for gathering and assessing citizens’ opinion on prospective policies. SUPER will use the +Spaces toolkit for deploying polling, debating and role-playing applications related to security and emergency management policies in different virtual spaces and for analyzing the users’ activities. This will allow policy makers better understand the roles that drive the security policy, the conditions that will maximize its impact but also to educate the citizens about the importance of security policy compliance and active citizen participation.
To make effective decisions using information derived from social media, it is important to observe the event from all possible perspectives, e.g. for political events, the perspective of those with left and right wing views. SUPER will develop technologies to automatically identify prominent communities present within social media during emergencies and incdidents and determine any bias and motivations they exhibit, exposing this information to key decision makers. Interaction networks within social media, e.g. follower/followee relationships in Twitter will be combined with state-of-the-art classification and clustering techniques to identify highly interconnected user groups and their commonalities.
One of the biggest challenges in social media content mining is the classifcation of sentimentally charged text which is associated with the ability to separate subjective from objective opinions and count the overall positive and negative feedback around a specific topic. In the context of SUPER we will develop a unified framework that will handle different types of social media (e.g. tweets, Facebook posts) and extract sentiment expressive patterns from user generated content. By studying sentiment diffusion in case of emergencies we will be able to estimate public opinion in security events.
Social media is often used to report upon ongoing emergencies in real-time from the scene. Indeed, it can be a valuable medium via which important incidents within a larger event can be reported to the emergency services. SUPER will investigate new approaches to automatically detect emerging incidents using social media as they occur, such that emergency responders can remain up-to-date as the situation changes over time. Large-scale distributed stream processing techniques will be developed in SUPER to detect new incidents relating to an event from bursty high-rate social streams.
What we are up to and where to meet
Papers from conferences and journals
Repors from the work we do
Want to know more?
Marco Cosentino (Project Coordinator)
Vitrociset S.p.A | B.U. Homeland Security
Via Tiburtina 1020
00156 Rome, Italy
Project full title: Social sensors for security assessments and proactive emergencies management
Grant agreement: N° 606853
Call identifier: FP7- Security - 2013.6.1-1
Total budget: 4,252,770.40 €
EU funding: 3,117,318.00 €
Duration: 01 April 2014 to 31 March 2017