Development of a Disaster Safety Sentiment Index via Social Media Mining

Abstract

People use social media platforms such as Twitter to record their personal thoughts and opinions. Social media platforms reflect people's sentiments as they are, and an accurate understanding of sentiments on social media could be useful and significant for disaster management. In this research, sentiment type modeling and sentiment quantification are proposed to understand the sentiments presented on social media platforms. Sentiment types are primarily analyzed based on the three major sentiments of affirmation, caution, and observation. Then, for a detailed understanding of sentiment progress according to the progress of a disaster or accident and the government's response, negative sentiments are categorized into anxiety, disappointment, depression, sadness, and displeasure to enhance the analysis, while positive sentiments are categorized into pleasure, happiness, and relief; Russell's circumplex model is used to develop a model of eight primary sentiments to acquire an overall understanding of the public's sentiments. Then, the sentiment index of each sentiment is quantified. Based on the results, the overall sentiment status of the public is monitored, and in the event of a disaster, the public's sentiment fluctuation rate can be quantitatively observed. Moreover, the influence of disasters and accidents on public sentiments, or the sentiment indices of different accidents, can be compared to identify the accidents that affect public sentiment and public needs after a disaster, and the insights can be used for policy-making.