Resumen de SEMA4A: a knowledge base for accessible evacuation and alert notifications in emergencies
Teresa Onorati
When an emergency occurs or is going to occur, the aim of organizations and agencies involved in the response phase is to restore quickly a safe situation and reduce the number of victims and damages. The notification of information about the kind of emergency, its characteristics, the location of safe places and available procedures for reaching them has a crucial role in order to facilitate the evacuation of citizens. Several organizations and agencies have been promoting the development of Information Technology (IT) tools, called Emergency Response Information Systems (ERIS) for the management of the activities performed in response to the emergencies. In particular, these systems provide modules for collecting, updating and notifying information about imminent disasters to potential affected people. Such notifications can be communicated through different channels, like websites, emails, text or voice messages. But to effectively inform people about an emergency, the notifications should be adapted automatically to each user’s profile (e.g. functional or contextual disabilities, elderly, children), the kind of emergency (e.g. typhoon, earthquake, tornado), the communication channel (e.g. PDAs, smartphones, pagers) and any other exceptional circumstances (e.g. interrupted roads, collapsed exit, dangerous area). For example, when a fire occurs in a building, a blind person should be alerted by audio signals or text messages (assuming she has a text-to-speech software on her device). Moreover, information can guide her to an assistant that can help her in reaching the exit. The efficacy of emergency notifications depends also on how different Emergency Notification Systems (ENS) communicate and interoperate with each other in order to share information even with different terminologies and types of disasters. For avoiding semantic incompatibilities, a common language is needed to improve the coordination not only among systems, but also among users. In fact, codifying the semantics of shared information in an accessible way could help citizens in interpreting notifications without misunderstandings and emergency operators in communicating among them. Modelling knowledge on alerting and evacuation processes, using expert systems, neural networks or ontologies, can help in personalizing emergency notifications and evacuation procedures. In particular, we posit that the knowledge base required for the personalization mechanism should cover at least four domains: accessibility, technology, emergency and evacuation procedures. These domains cover the factors to take into account for adapting the notifications. Consequently, the accessibility is considered a representation for the user’s profile, technology for the interactive devices and the communication channel, emergency for the characteristics of the situation and evacuation procedures for the escaping measures. In this thesis, we propose the design of an ontology called SEMA4A (Simple Emergency Alerts 4 [for] All). The ontology is a knowledge representation based on semantic rules that allows to model articulated knowledge through the definition of complex relations among concepts from different domains. This choice is also related to the possibility of using specific tools based on first order logic for verifying the validity and the integrity of the proposed representation. The development of the ontology has to meet the objectives that motivated this research work: consistency, completeness, understandability and interoperability with existent systems and protocols. For the consistency, we have run a reasoner tool called Pellet obtaining that there are not redundancies and the mapping is syntactically coherent. Concerning completeness and understandability, we have performed a quantitative and a qualitative evaluation. The goal of the quantitative evaluation is to compute three well-known functions in the domain of ontological engineering: precision, coverage and accuracy. These three measures evaluate how much the ontology is representative respect to the domains of interest (i.e. accessibility, emergency, evacuation and technology). In the qualitative evaluation, we have involved international experts in accessibility, evacuation and emergency to test the validity of the proposed mapping with respect to their expertise. Finally, the interoperability has been guaranteed codifying SEMA4A with a standard language called OWL (Ontology Web Language) and following formal recommendations published as an initiative of the W3C (World Wide Web Consortium). Taking into account the results obtained from the evaluations, we posit that the proposed ontology addresses needed information for sharing and integrating alert notifications about emergencies and evacuation procedures into existent solutions (i.e. notification mechanisms, information systems, communication protocols). As proof of this, we have developed three use cases in collaboration with the DEI Group of the University Carlos III of Madrid. SEMA4A has been applied for adapting available information considering several factors: the user’s profile, the kind of emergency, the communication channel and other exceptional circumstances. The first use case, called CAPONES, sends emergency alerts adapting the content and the visualization to the needs of involved users. The second system is NERES which aims at generating and notifying personalized evacuation routes. The last case is the EmergenSYS platform that provides three different mobile tools for sending alerts in two directions: from citizens to emergency operators and from emergency operators to citizens. -----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
