Abstract: In this paper, we introduce Tweedr, a Twitter-mining tool that extracts actionable information for disaster relief workers during natural disasters. The Tweedr pipeline consists of three main parts: classification, clustering and extraction. In the classification phase, we use a variety of classification methods (sLDA, SVM, and logistic regression) to identify tweets reporting damage or casualties. In the clustering phase, we use filters to merge tweets that are similar to one another; and finally, in the extraction phase, we extract tokens and phrases that report specific information about different classes of infrastructure damage, damage types, and casualties. We empirically validate our approach with tweets collected from 12 different crises in the United States since 2006.

Abstract: Disaster risk management practitioners have the responsibility to make decisions at every phase of the disaster risk management cycle: mitigation, preparedness, response and recovery. The decisions they make affect human life. In this paper, we consider the current state of the use of AI in information extraction (IE) for disaster risk management (DRM), which makes it possible to leverage disaster information in social media. We consolidate the challenges and concerns of using AI for DRM into three main areas: limitations of DRM data, limitations of AI modeling and DRM domain-specific concerns, i.e., bias, privacy and security, transparency and accountability, and hype and inflated expectations. Then, we present a systematic discussion of how explainable AI (XAI) can address the challenges and concerns of using AI for IE in DRM.

Abstract: Networking in the event of disasters requires new hybrid wireless architectures such as Wireless Mesh Networks (WMNs). Provisioning Quality of Service (QoS) in such networks which are quickly deployed during emergencies demand radical solutions. In this paper, we provide a new QoS approach for voice calls over a wireless mesh networks during emergency situations. According to our scheme, the contention and back-off parameters are modified based on the emotion content in the voice streams. This paper also looks at methods for detecting emotion from an incoming voice call using the speech signal. The issues of interest in such situations are whether the caller is in a state of extreme panic, moderate panic, or in a normal state of behavior. The communication network behavior should be modified to provide differentiated QoS for calls based on the degree of emotion. We use several features extracted from the speech signal like the range of pitch variation, energy in the critical bark band, range of the first three formant variations, and speaking rate among others to discriminate between the three emotional states. At the back end the Gaussian mixture modeling techniques is used to model the three emotional states of the speaker. Since a large number of features increase the computational complexity and time, a feature selection technique is employed based on the Bhattacharya distance, to select the set of features that give maximum discrimination between the classes. These set of features are employed to simulate an emotion recognition system. The results indicate a promising emotion detection rate for the three emotions. We also present the early results on detecting the emotion content in the speech and using this in the MAC layer differentiated QoS provisioning scheme. Our scheme provides an end-to-end delay performance improvement for panicked calls as high as 60% compared to normal calls.

Abstract: Social media plays a vital role as a communication source during large-scale disasters. The unstructured and informal nature of such short individual posts makes it difficult to extract useful information, often due to a lack of additional context. The potential of social media threads– sequences of posts– has not been explored as a source of adding context and more information to the initiating post. In this research, we explored Twitter threads as an information source and developed an information extraction model capable of extracting relevant information from threads posted during disasters. We used a crowdsourcing platform to determine whether a thread adds more information to the initial tweet and defined disaster-related information present in these threads into six themes– event reporting, location, time, intensity, casualty and damage reports, and help calls. For these themes, we created the respective thematic lexicons from WordNet. Moreover, we developed and compared four information extraction models trained on GloVe, word2vec, bag-of-words, and thematic bag-of-words to extract and summarize the most critical information from the threads. Our results reveal that 70 percent of all threads add information to the initiating post for various disaster-related themes. Furthermore, the thematic bag-of-words information extraction model outperforms the other algorithms and models for preserving the highest number of disaster-related themes.

Abstract: This paper presents our work on developing a system for Internet monitoring and knowledge extraction from different web documents which contain information about disasters. This system is based on ontology of the disasters domain for the knowledge extraction and it presents all the information extracted according to the kind of the disaster defined in the ontology. The system disseminates the information extracted (as a synthesis of the web documents) to the users after a filtering based on their profiles. The profile of a user is updated automatically by interactively taking into account his feedback.

Abstract: The design of a WSN for environmental data monitoring is a largely ad-hoc human process. In this paper, we propose the automatic generation of queries for sensor data extraction, based on the collection of a number of parameters concerning the physical phenomenon to be controlled, the relevant physical variables, the types of sensors to be deployed and their allocation, the data collection frequencies, and other features.

Abstract: Today's crises attract great attention on social media, from local and distant citizens as well as from news media. This study investigates the possibilities of real-time and automated analysis of Twitter messages during crises. The analysis was performed through application of an information extraction tool to nearly 97,000 tweets that were published shortly before, during and after a storm hit the Pukkelpop 2011 festival in Belgium. As soon as the storm hit the festival tweet activity increased exponentially, peaking at 576 tweets per minute. The extraction tool enabled analyzing tweets through predefined (geo)graphical displays, message content filters (damage, casualties) and tweet type filters (e.g., retweets). Important topics that emerged were 'early warning tweets', 'rumors' and the 'self-organization of disaster relief' on Twitter. Results indicate that automated filtering of information provides valuable information for operational response and crisis communication. Steps for further research are discussed. © 2012 ISCRAM.

Abstract: As crisis are complex systems, providing an accurate response to an ongoing crisis is not possible without ensuring situational awareness. The ongoing works around knowledge management and ontologies provide relevant and machine readable structures towards situational awareness and context understanding. Many metamodels, that can be derived into ontologies, supporting the collect and organization of crucial information for Decision Support Systems have been designed and are now used on specific cases. The next challenge into crisis management is to provide tools that can process an automated population of these metamodels/ontologies. The aim of this paper is to present a strategy to extract concept-instance relations in order to feed crisis management ontologies. The presented system is based on a previously proposed generic metamodel for information extraction and is applied in this paper to three different case studies representing three different crisis namely Ebola sanitarian crisis, Fukushima nuclear crisis and Hurricane Katrina natural disaster.