An Empirical Methodology for Detecting and Prioritizing Needs during Crisis Events

Abstract

In times of crisis, identifying essential needs is crucial to providing appropriate resources and services to affected entities. Social media platforms such as Twitter contain a vast amount of information about the general public's needs. However, the sparsity of information and the amount of noisy content present a challenge for practitioners to effectively identify relevant information on these platforms. This study proposes two novel methods for two needs detection tasks: 1) extracting a list of needed resources, such as masks and ventilators, and 2) detecting sentences that specify who-needswhat resources (e.g., we need testing). We evaluate our methods on a set of tweets about the COVID-19 crisis. For extracting a list of needs, we compare our results against two official lists of resources, achieving 0.64 precision. For detecting who-needs-what sentences, we compared our results against a set of 1,000 annotated tweets and achieved a 0.68 F1-score.

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

“…(Extractive) text summarization identifies the most salient sentences in a corpus (Eisenstein 2019). Needs detection identifies the resources needed during a disaster, e.g., by extracting tweets that mention resources needed (Basu et al 2017) or creating a ranked list of resources (Sarol et al 2020).…”

Question Answering techniques for detecting needs of people in crisis

“…During crisis, not only affected people wrote for help also the people who can help and give any available to affected people. Therefore, A semi-automated platform called NARMADA is proposed to collect all social media posts which include all information about assisting post-disaster relief coordination efforts by applying Natural Language Processing and Information Retrieval techniques for detecting the needs and the available need-related resources to match the needs and its suitable resource [12], or detect the needs from social media but using the resources from the official organization and agent like World Health Organization's (WHO) in case of COVID-19 which provide resource planning guidelines by proposing two methods for 2 distinct but related needs detection tasks (1)detect the top needs for COVID19 by using word embedding to obtain the closest terms to needs and supplies(2) Specific needs detection to identify people who needs particular resource by using rule-based methodology [14]. Most of previous related work focus for detecting other details related to crisis not only the type of crisis but also the related sub events and the needs of affected people which is the major purpose in the crisis management process.…”

Detection and Categorization of Needs during Crises Based on Twitter Data

“…However, prior research has shown that at least in some situations, it is important to distinguish between different types of crises, especially man-made versus natural ones, as they require different response strategies to effectively address unique characteristics and challenges (Alexander 2005;Palen et al 2009). Natural crises such as hurricanes, earthquakes, and floods are often unpredictable and require immediate action to save lives and minimize damage (Sarol et al 2020;Sarol, Dinh, and Diesner 2021). Man-made crises such as terrorist attacks or warfare often involve intentional harm or negligence and require approaches involving intelligence, law enforcement, security measures, and efforts to ensure the safety of affected populations (Chang, Diesner, and Carley 2012;Diesner and Carley 2010;Petrescu-Prahova and Butts 2005;Tierney and Trainor 2004).…”

“…Prior literature in crisis informatics highlights the need for understanding the distinct characteristics and dynamics of different types of crises, including their causes, impact, and response requirements (Olteanu, Vieweg, and Castillo 2015;Sarol et al 2020). Detecting needs related to a crisis, a task driven by advances in natural language processing (NLP), has shown to be effective in reliably identifying specific challenges associated with different crisis events (e.g., and types (e.g., biological disaster) (Sarol et al 2020;Basu et al 2017;Sarol, Dinh, and Diesner 2021).…”

“…In this paper, we analyze (utterances involving) mentions of needs and their text-based contextualization based on tweets, not needs that people or groups may actually have per se. Prior work on systematically detecting needs using NLP methods has used text summarization (Sarol, Dinh, and Diesner 2021;Rudra et al 2015), word embeddings (Sarol et al 2020;Basu et al 2017), and classification using linguistic features (Olteanu, Vieweg, and Castillo 2015;Verma et al 2011). For example, Sarol, Dinh, and Diesner (2021) applied extractive text summarization with termfrequency and keyword-in-context scores as word-level features to crisis texts, and extracted the most salient sentences that were crisis-related.…”