Visual Analytics of Microblog Data for Pandemic and Crisis Analysis

Microblogging is a rich and plentiful source of data that contains potentially valuable information amidst noise. This work presents software that extracts useful metadata from a microblog dataset to explore and analyze the data for the detection and exploration of crisis events. The developed software (Vambutu) was successfully used to examine an artificially-generated dataset for the onset and source of an illness outbreak. A part of speech tagger was used to divide microblog posts into their component parts for the purposes of identifying posts pertaining to first, second, and third-hand experiences. A successful demonstration of this ability revealed clearly identifiable patterns for first-hand experiences. For example, for the word pneumonia we found patterns that were not apparent when all posts pertaining to pneumonia were examined at once. This promising result demonstrates the potential as a tool for filtering out irrelevant noise during the occurrence of a crisis event.