Researchers at the University of Guelph have developed a system to help track the spread of infectious diseases using Twitter, a framework they say has the potential to identify outbreaks faster than official monitoring agencies.
The study, which tracked global outbreaks of avian flu, may be the first of its kind to use the social network as an early warning system for tracking diseases.
The team of veterinary and computer science researchers used artificial intelligence and machine learning to analyze over 200,000 tweets for mentions of avian flu, and then sorted that data to help reveal clues about where the disease is spreading.
They say they detected 75 per cent of real-world outbreaks of avian flu in poultry between between July 2017 and November 2018.
“Our web robot monitors the Internet 24/7,” said Shayan Sharif, a professor and associate dean at the Ontario Veterinary College. “We then go back to those tweets, we try to annotate them essentially to identify a background noise from the actual data that can be utilized in order to train the machine.”
Potential use against coronavirus
They’re now turning their attention to COVID-19 as well.
“In the context of coronavirus, people may talk about their signs and symptoms such as fevers, aching all over their body, or having shortness of breath,” he said. “We’d then try and utilize those types of keywords in order to train our machine.”
Sharif said the framework has been able to identify potential avian flu outbreaks before conventional monitoring systems, something he hopes could be of help monitoring COVID-19 and other human pathogens.
He said, the immediacy of Twitter data means monitors can glean information on a potential outbreak without having to wait for results from distant laboratories.
“There’s is a huge gap between the time that a case is detected and the time that the case is confirmed, for example, coronavirus or avian influenza,” he said. “People start talking about their own signs and symptoms well ahead of that.”
The research was published in the journal Nature Scientific Reports in December. The paper’s lead author, computer science professor Rozita Dara, says they hope the system can be used to predict outbreaks earlier “as a supplemental source of surveillance.”