CSIRO monitoring Twitter to detect disasters
CSIRO monitoring Twitter to detect disasters
04 May 2014
published by www.smh.com.au
Australia — Twitter could be faster at detecting earthquakes, bushfires, floods, cyclones and even terrorist attacks than conventional emergency sources with new CSIRO software monitoring tweets and automatically alerting emergency services.
The Emergency Situation Awareness tool operates around the clock analysing millions of tweets and takes about three minutes to detect an earthquake.
The system has reported 22 earthquakes for Australia and New Zealand this year.
Dr Mark Cameron, from CSIRO, said; what people see, hear and do before, during and after a crisis or emergency is reported. The challenge for emergency service organisations is to harness community intelligence about unfolding situations in order to direct and fine-tune their communications, he said.
Advertisement
He said the system not only identifies an emergency event has occurred, but can also provide information on the impact of the event on infrastructure such as roads, bridges, power and telecommunications and report cries for help.
The CSIRO is looking to extend its twitter monitoring alert tool to include south-east Asian partners.
A prototype of the Twitter filter was used by the CSIRO to monitor the Qantas Flight 32 engine failure in November 2010 and the Grantham-Toowoomba flash floods in January 2011.
The system was developed further and has been used by the federal governments crisis co-ordination centre, the joint Australian tsunami warning centre at Geoscience Australia and by the New South Wales fire authorities including during the bushfires of October 2013.
Law enforcement agencies are also believed to be interested in the Twitter monitoring system.
The system monitors the statistical incidence of words used by people on Twitter to describe emergency events and compares this with historical word occurrences on Twitter from past disaster events.
The high use of words such as bushfire and earthquake in a rapid sequence is called a burst detection.
[The tool] counts the number of times a word appears in a sliding five-minute window across the Twitter stream. If this is significantly greater than expected, an alert is generated, the organisation said. An alert is then emailed to emergency agencies.
Dr Cameron said the event is then confirmed by an expert before a warning is issued to the public. Its very important to have a human in the loop to verify an alert and then translate that alert into a communication about the actions an impacted community can or should take, he said.
