The role of social media in measuring human response to urban flash flooding

Abstract

This thesis explores the role of social media in urban flooding. The author analyzes the activity of weibos (Chinese tweets) related to Beijing's "7.21" flash flood in the Sinaweibo system (the most popular social media open platform in China) and characterizes these weibos from the 37 hours following this disaster. In order to understand the response of the public to urban flash flooding, multiple methods are used, including trend analysis, content analysis for high frequent terms and co-occurrence words, and lexicon-based sentiment analysis. In particular, weibos with geo-location information were extracted to draw different sentiment maps for the city. Sentiment maps show the public emotion (polarity, intensity and type) geographically. Through these analyses, I set out to construct a framework to process massive amount of data generated by social media, and proposed a methodology for converting the data into actionable knowledge. This work explored extracting emotions from weibos, distilling crowd-wisdom by using filters and algorithms to smooth out the noise in the massive amount of data, and determining that human emotions closely correlated with severe natural disaster. By tracking human emotions, it was possible to track the progress of the disaster, and more importantly whether relief was provided to mitigate the disaster. Moreover, by projecting the emotional polarity, intensity and emotional type onto maps, the visualization can provide reasonably clear and timely picture of when and where the strongest emotions occurred. The methodology developed in this thesis could facilitate innovative approaches in the field of urban disaster planning.