Twitter may be just a microblogging platform, but some scientists believe they can use tweet patterns across the site to instead predict natural phenomena.
Researchers from the University of Dundee are using artificial intelligence to comb through Twitter, in an effort to develop an early-warning system for flood-prone communities.
Dr Roger Wang and his colleagues from the university’s Science and Engineering department are trying to show how AI can be used to extract patterns from the readily available crowdsourced data on Twitter and other apps to make predictions and monitoring systems for urban flooding.
It’s hard to capture data on flooding in cities, which in turn makes it tough to conduct risk analysis and plan control measures in detail. So, instead, the Dundee researchers are gathering hyperlocal data from social media apps to complement datasets based on traditional remote sensing and witness reports. So far, their efforts actually seem to be working.
“Sea levels have been rising at an average rate of 3.4mm a year over the past decade,” Dr Wang says. “The extremes of today will become the average of the future so coastal cities and countries must take action to protect their land.”
“A tweet can be very informative in terms of flooding data. Key words were our first filter, then we used natural language processing to find out more about severity, location and other information. Computer vision techniques were applied to the data collected from MyCoast, a crowdsourcing app, to automatically identify scenes of flooding from the images that users post.”
The streamed Twitter data for a month in 2015, using filter keywords like “flood”, “inundation”, “dam”, “dike”, and “levee”. More than 7500 tweets were analysed in the 30 days. That information was confirmed against precipitation data and road closure reports. The scientists found that flood-related tweets more common at higher precipitation levels, also matching with the road closure reports.
“We have reached the point of 70 percent accuracy,” Dr Wang continues. “This can be then used to improve forecasting models and early warning systems to help residents and authorities prepare for an upcoming flood.”