#pray4victims: Consistencies in Response to Disaster on Twitter

This paper studies commonalities in response across disasters in online social networks (OSNs) and Twitter specifically. After presenting an algorithm for extracting vocabularies across disasters, we extract type-specific vocabularies for terrorist attacks, earthquakes, and climate-related disasters between 2012 and 2017. Within similar disasters, commonalities emerge: terrorism responses reference the "attack" and law enforcement, earthquake responses mention the quake and its magnitude, and climate-related responses include safety and requests for aid. Across disaster types, tweets regularly mention victims/affected and prayer, consistent with communal coping and social support in crisis aftermath. Using these disaster-type vocabularies, we study Twitter as an alternate measure for severity, correlating casualties to Twitter volume. These vocabularies better correlate with casualties than baseline crisis lexica, especially in western countries. Twitter response and casualties diverge at the extreme, and Twitter response is stronger in Western countries, suggesting perceived severity is driven by additional factors.