Developing a Sina Weibo Incident Monitor for Disasters

Developing a Sina Weibo Incident Monitor for Disasters Bella Robinson* bella.robinson@csiro.au Hua Bai*,** hua.bai@csiro.au Robert Power* robert.power@csiro.au Xunguo Lin* xunguo.lin@csiro.au * CSIRO Digital Productivity Flagship G.P.O. Box 664 Canberra, ACT 2601, Australia ** School of Management Harbin Institute of Technology 92 Xidazhi Street, Harbin, Heilongjiang, China, 150001 Abstract This paper presents ongoing work to de- velop an earthquake detector based on near-real-time microblog messages from China. The system filters earthquake re- lated keywords from Sina Weibo messages available on the public timeline and uses a classifier to determine if the messages cor- respond to people experiencing an earth- quake. We describe how the classifier has been established and report preliminary results of successfully using it as a detec- tor for earthquake events in China. We also provide an overview of how we have accessed messages in Chinese from Sina Weibo, including a summary of their structure and content. We note our expe- rience of processing this text with Natu- ral Language Processing packages and de- scribe a preliminary web site for users to view the processed messages. Our long term aim is to develop a general alert and monitoring system for various disaster event types in China reported by the public on Sina Weibo. This first case study provides a working example from which an ‘all hazards’ system can be de- veloped over time. 1 Introduction Natural disasters are a major cause of loss and damage to both lives and properties around the world. Many disasters impact regions and popu- lations with little warning which is made worse by the projected impacts of climate change and in- creasing urbanization of the world’s population. For these reasons, it is important to enhance the ability and effectiveness of emergency manage- ment. One of the pressing challenges while managing an emergency or crisis event is the collection and communication of reliable, relevant and up to date information. Timely, accurate and effective mes- sages play a vital role for disaster response. For emergencies, a rapid response is needed to make effective decisions and to mitigate serious damage. Losses can be reduced by providing information to both rescue organizations and potential victims. Microblogging services have proven to be a use- ful source of information to emergency managers to gain first hand information about disaster events from the community experiencing them (Abel et al., 2012; Cameron et al., 2012; Chowdhury et al., 2013; Zhou et al., 2013). This new source of in- formation can be used by emergency coordinators and responders to provide appropriate services to the affected area. It can also be used by the wider community to seek timely information about the event and as a means of engaging with the unfold- ing situation from a safe distance. Work to date has focused predominantly on Twitter as the social media microblogging source. This service is not available in China and we wanted to explore processing techniques using messages from Sina Weibo, the Chinese equiva- lent to Twitter. Our initial task is to identify earth- …

Content truncated to 500 words