Natural Language Processing to the Rescue? Extracting "Situational Awareness" Tweets During Mass Emergency

Abstract

In times of mass emergency, vast amounts of data are generated via computer-mediated communication (CMC) that are difficult to manually cull and organize into a coherent picture. Yet valuable information is broadcast, and can provide useful insight into time- and safety-critical situations if captured and analyzed properly and rapidly. We describe an approach for automatically identifying messages communicated via Twitter that contribute to situational awareness, and explain why it is beneficial for those seeking information during mass emergencies.We collected Twitter messages from four different crisis events of varying nature and magnitude and built a classifier to automatically detect messages that may contribute to situational awareness, utilizing a combination of hand-annotated and automatically-extracted linguistic features. Our system was able to achieve over 80% accuracy on categorizing tweets that contribute to situational awareness. Additionally, we show that a classifier developed for a specific emergency event performs well on similar events. The results are promising, and have the potential to aid the general public in culling and analyzing information communicated during times of mass emergency.

Variation in Situational Awareness Information due to Selection of Data Source, Summarization Method, and Method Implementation

“…While data gathered from online sources have been shown to aid emergency responders in constructing SI (Verma et al 2011;Olteanu, Vieweg, and Castillo 2015), they may entail misinformation (Castillo, Mendoza, and Poblete 2011), false rumors (Starbird et al 2014), polarization of opinions, and echo chambers (Barberá et al 2015), all of which may confound response efforts. We also found that misinformation can seep into the summary sentences.…”

“…This represents a shift in main sources for gaining SI from television, radio, and print media to real-time or near-time updates, e.g., via Twitter, blogs, and online news. Researchers in crisis informatics (Palen and Anderson 2016;Reuter, Hughes, and Kaufhold 2018) have sought computational means to detect and examine crisis-related information from different text-based sources (Verma et al 2011;Abel et al 2011;Tkachenko, Jarvis, and Procter 2017) in a timely and unbiased manner. Using Twitter as a primary source, Verma and colleagues (2011) extracted linguistic features (unigrams, bigrams, parts-of-speech, subjective cues, register, tone) to detect tweets that contain content related to SI.…”

“…In a post hoc analysis of the U.S.' response to the Haiti earthquake, the deployed responders did not have full SI before they arrived in Haiti, which impacted planning and coordination (Cecchine et al 2013). To mitigate information overload and quickly obtain SI, a number of studies has focused on automatically (e.g., via machine learning (Imran, Mitra, and Castillo 2016)) extracting relevant information (Weil et al 2008), such as alerts about dangerous situations, missing people, and building damages, from sources such as Twitter (Verma et al 2011), blogs and online forums (Li and Chen 2008;Kitamoto 2005), and news articles (Tanev, Zavarella, and Steinberger 2017). In this paper, we refer to information related to situational awareness about an event and extracted from text data as situational awareness information (SII), which is different from raw data or information about crisis events.…”

A Multi-label Classification of Disaster-Related Tweets with Enhanced Word Embedding Ensemble Convolutional Neural Network Model

“…They studied the behavior of microbloggers by conducting a qualitative analysis of tweets published during a flooding incident. The authors of [11] employed the bags-ofwords method to locate a crisis's data. Various grammatical elements, like Parts-Of-Speech (POS), are influenced by the vocabulary used in Twitter posts.…”

Enhancing Event Descriptions through Twitter Mining

“…We utilise some NLP techniques in the process of query creation and tweet relevance ranking. Other authors report on techniques for classification of 'event' and 'non-event' tweets that deploy shallow linguistic analysis, e.g., (Verma et al 2011) reports on an approach of using linguistic features (e.g., objectivity, impersonality, formality, etc.) for detecting tweets with content relevant to situational awareness during mass emergencies.…”