Abstract: Over the years, with the increase in social media engagement, there has been an in increase in various pipelines to analyze, classify and prioritize crisis-related data on various social media platforms. These pipelines utilize various data augmentation methods to counter imbalanced crisis data, sophisticated and off-the-shelf models for training. However, there is a lack of comprehensive study which compares these methods for the various sections of a pipeline. In this study, we split a general crisis-related pipeline into 3 major sections, namely, data augmentation, model selection, and training methodology. We compare various methods for each of these sections and then present a comprehensive evaluation of which section to prioritize based on the results from various pipelines. We compare our results against two separate tasks, information classification and priority scoring for crisis-related tweets. Our results suggest that data augmentation, in general,improves the performance. However, sophisticated, state-of-the-art language models like DeBERTa only show performance gain in information classification tasks, and models like RoBERTa tend to show a consistent performance increase over our presented baseline consisting of BERT. We also show that, though training two separate task-specific BERT models does show better performance than one BERT model with multi-task learning methodology over an imbalanced dataset, multi-task learning does improve performance for more sophisticated model like DeBERTa with a much more balanced dataset after augmentation.

Abstract: We developed a web application for location recording and rescue request using Twitter (T-Pl@ce). This application helps supported users (e.g., older adults, persons with disabilities, and children) who require support to share their location coordinates via Twitter. Supporting users (e.g., families, relatives, or neighbors) of the supported user can then check the location coordinates of the supported user when required. When the supported user needs to be rescued, he/she can post a rescue request on Twitter by pressing the “Rescue request” button on the application. In this study, we introduce the e-mail notification function to reliably notify a rescue request to the system administrator. In addition, to track the location of the supported user, we introduce a location tracking function. Then, the administrator, the emergency assistance employees (e.g., rescue experts or social workers), or the supporting user can refer to the request and the location tracking page and execute the support and rescue activities.

Abstract: Crowdsourcing is an increasingly popular approach for processing data in response to disasters. While volunteer crowdsourcing may suÿce for high-profile disasters, paid crowdsourcing may be necessary to recruit workers for less prominent events. Thus, understanding the impact of payment schemes on worker behavior and motivation may improve outcomes. In this work, we presented workers recruited from Amazon Mechanical Turk with a disaster response task in which they could provide a variable number of image ratings. We paid workers a fixed amount to provide a minimum number of image ratings, allowing them to voluntarily provide more if desired; this allowed us to examine the impact of dierent amounts of required work. We found that requiring no ratings resulted in workers voluntary completing more work, and being more likely to indicate motivation related to interest on a post survey, than when small numbers of ratings were required. This is consistent with the motivational crowding-out eect, even in paid crowdsourcing. We additionally found that providing feedback on progress positively impacted the amount of work done.

Abstract: Crowdsourcing is used to enlist workers as a resource for a variety of applications, including disaster response. However, simple tasks such as image labeling often feel monotonous and lead to worker disengagement. This provides a challenge for designing successful crowdsourcing systems. Existing research in the design of work indicates that task variety is a key factor in worker motivation. Therefore, we asked Amazon Mechanical Turk workers to complete a series of disaster response related subtasks, consisting of either image labeling or locating photographed areas on a map. We varied the frequency at which workers encountered the dierent subtask types, and found that switching subtask type at dierent frequencies impacted measures of worker engagement. This indicates that a certain amount of variety in subtasks may engage crowdsourcing workers better than uniform subtask types.

Abstract: We address the problem of making sense of rumour evolution during crises and emergencies. We study how
understanding and capturing emerging rumours can benefit decision makers during such event. To this end, we
propose a two-step framework for detecting rumours during crises. In the first step, we introduce an algorithm to
identify noteworthy sub-events in real time. In the second step, we introduce a graph-based text ranking method
for summarising newsworthy sub-events while events unfold. We use temporal and content-based features to
achieve the effective and real-time response and management of crises situations. These features can improve
efficiency in the detection of key rumours in the context of a real-world application. The effectiveness of our
method is evaluated over large-scale Twitter data related to real-world crises. The results show that our framework
can efficiently and effectively capture key rumours circulated during natural and human-made disasters.

Abstract: We present Tweet4act, a system to detect and classify crisis-related messages communicated over a microblogging platform. Our system relies on extracting content features from each message. These features and the use of an incident-specific dictionary allow us to determine the period type of an incident that each message belongs to. The period types are: Pre-incident (messages talking about prevention, mitigation, and preparedness), during-incident (messages sent while the incident is taking place), and post-incident (messages related to the response, recovery, and reconstruction). We show that our detection method can effectively identify incident-related messages with high precision and recall, and that our incident-period classification method outperforms standard machine learning classification methods.

Abstract: Social Media contain a wealth of information that could improve the situational awareness of Emergency Managers during a crisis, but many barriers stand in the way. These include information overload, making it impossible to deal with the flood of raw posts, and lack of trust in unverified crowdsourced data. The purpose of this project is to build a communications bridge between emergency responders and technologists who can provide the advances needed to realize social media?s full potential. We are employing a Delphi study survey design, which is a technique for exploring and developing consensus among a group of experts around a particular topic. Participants include emergency managers and technologists with experience in software to support the use of social media in crisis response, from many countries. The topics of the study are described and preliminary, partial results presented for Round 1 of the study, based on 33 responses.

Abstract: Several recent studies point the way to enabling emergency response managers to be able to find relevant posts and incorporate them into their sensemaking and decision making processes. Among the approaches that have improved the ability to find the most relevant information are the social conventions of creating topic groups and tags and of “retweeting;” the use of trained volunteers to filter and summarize posts for responders; automated notifications of trending topics; natural language processing of posts; techniques for identifying posts from the disaster site; and the use of GIS and crisis maps to visually represent the distribution of incidents.

Abstract: During large-scale disasters it is not uncommon for Public Safety Answering Points (e.g., 9-1-1) to encounter
service disruptions or become overloaded due to call volume. As observed in the two past United States hurricane
seasons, citizens are increasingly turning to social media whether as a consequence of their inability to reach
9-1-1, or as a preferential means of communications. Relying on past research that has examined social media
use in disasters, combined with the practical knowledge of the first-hand disaster response experiences, this paper
develops a knowledge-driven framework containing parameters useful in identifying patterns of shared
information on social media when citizens need help. This effort explores the feasibility of determining
differences, similarities, common themes, and time-specific discoveries of social media calls for help associated
with hurricane evacuations. At a future date, validation of this framework will be demonstrated using datasets
from multiple disasters. The results will lead to recommendations on how the framework can be modified to make
it applicable as a generic disaster-type characterization tool.

Abstract: This paper presents a case study of a very recent man-made crisis in Norway on 22 July, 2011, during which a single person first detonated a bomb in downtown Oslo and then killed 69 young people on the island of Utøya. It proposes a novel way of conceptualizing the public contribution to mobilization of resources using microblogging, particularly tweeting. By examining aspects of public and professional response to this crisis, the notion of peripheral response is developed in relation to emergent forms of agile and dialogic emergency response. Through examining the distributed efforts of responding to the crisis, the paper also revisits situation awareness and reflects upon the dynamic and constantly changing environment that social media and crises inhabit together. © 2012 ISCRAM.

Abstract: The recent rise of Social Media services has created huge streams of information which can be very valuable in a variety of scenarios. One specific scenario that has received interest is how Social Media analytics can be beneficial in crisis situations. In this paper, we describe our vision for a Social Media-ready command and control center. As motivation for our work, we present a short analysis of tweets issued in NYC during Hurricane Sandy in late October 2012 and we give an overview of the architecture of our event detection subsystem.

Abstract: Twitter is always worthwhile in facilitating communication during disasters. It helps in raising situational awareness and undertaking disaster control actions as quickly as possible to alleviate the miseries. But the noisy essence of Twitter causes difficulty in distinguishing relevant information from the heterogeneous contents. Therefore, extracting informative tweets is a substantial task to help in crisis intervention. Analyzing only the text or image content of the tweet often misses necessary insights which might be helpful during disasters. In this paper, we propose a multimodal framework to address the challenges of identifying informative crisis-related tweets containing both texts and images. Our presented approach incorporates an early fusion strategy of BERT-LSTM and ResNet50 networks which effectively learns from the joint representation of texts and images. The experiments and evaluation on the benchmark CrisisMMD dataset show that our fusion method surpasses the baseline by 7% and substantiates its potency over the unimodal systems.

Abstract: Since the beginning of this millennium, there has been an increasing use of social media and crowdsourcing (SMCS) technologies in disaster situations (Reuter & Kaufhold, 2018). Disaster management organizations and corresponding research are increasingly working on ways of integrating SMCS into the processes of crisis management. In a changing technological landscape to address disasters, and with increasing diversity of stakeholders in disasters, the purpose of this research is to provide an overview of technologies for SMCS within disasters to improve community resilience. The identified and analyzed technologies are summarized under the term “Disaster Community Technologies” (DCT). The paper presents a classification schema (the “DCT-schema”) for those technologies. The goal is to generate an overview of DCT in a rapidly evolving environment and to provide the practical benefit for different stakeholders to identify the right one from the overview.