Sentiment Analyses of Twitter for Winter Storm Leo

Introduction

Natural disasters¹ have been in increasing events in our contemporary era. According to the United Nations Office for Disaster Risk Reduction (UNDRR) (2020), 6,873 natural disasters occur worldwide, which affect 510,837 deaths and 3.9 billion people between 2000 and 2019. In 2019, 396 natural disasters are recorded in the Emergency Events Database (EM-DAT) with 11,755 deaths, 95 million people affected and 103 billion US$1in economic losses across the world (EM-DAT, 2020).

Prior studies have heavily highlighted how natural disasters play an important role in negative sentiment (This study defines human sentiment as an emotion or attitude toward an event.) (see e.g., Beaglehole et al., 2018; Baral & Bhagawati, 2019; Bernstein & Pfefferbaum, 2018; Brown et al., 2017; McGuire et al., 2018). For example, Beaglehole et al., (2018) report that continuous measures of psychological distress are increased after natural disasters based on systematic review and meta-analysis for 41articles. Baral and Bhagawati (2019) show that post traumatic stress disorder (PTSD) are prevalent among 24.10% of adult survivors with highest intrusion symptoms based on a sample of 291 adult survivors after 10 months of Nepal Earthquake 2015.

Some scholars have reported how natural disasters exert some impacts on positive sentiment, whereas they have shown the negative impacts of natural disasters more than the positive impacts of those (see e.g., Buscaldi & Hernandez-Farias, 2015; Pourebrahim et al., 2019; Shalunts et al., 2014). For instance, Buscaldi and Hernandez-Farias (2015) report 499 tweets (3.7%) as positive, 3519 tweets (26%) as negative, 1019 tweets (7.5%) as ironic, and 8922 tweets (65.9%) as subjective for the 2014 Genoa Floodings. Shalunts et al. (2014) show that a few selected accounts exhibit higher ratios of positive sentiment, whereas negative sentiment dominates among the 50 most frequent Twitter users in the dataset of German during the Central European floods of 2013.

Scholars have especially utilized Social Network Systems (SNS), such as Twitter and Facebook, with the development of Information Technology (IT) after the 2010s for understanding the effects of natural disasters on human sentiment to provide new implications (see e.g., Karami et al., 2020; Reynard & Shirgaokar, 2019; Zou et al., 2018). For instance, Zou et al. (2018) show the spatial–temporal patterns of Twitter activities in 126 counties during Hurricane Sandy for improved understanding of disaster resilience. Reynard and Shirgaokar (2019) employ Twitter data to understand allocation decisions during Hurricane Irma in 2017 in Florida. Karami et al. (2020) highlight the negative concerns of people during the 2015 South Carolina flood by employing Twitter situational awareness.

However, many studies have not considered a magnitude of the importance of contents and a magnitude of different regions and periods for the relationship between natural disasters and human sentiment (see e.g., Alexander, 2010; Barnett, 1999; Bhavaraju et al., 2019; Bruns & Stieglitz, 2012; Burby, 2006; Healy & Malhotra, 2009; Rivera & Miller, 2006; Wang & Taylor, 2016). The magnitude of the importance of information and a magnitude of different regions and periods would exert a significant impact on human sentiment for natural disasters.