With the increase in the number of fatal cases associated with the coronavirus, a group of researchers decided to apply machine learning techniques on social networks to detect signs of spread of the disease. The virus appeared in Wuhan, China, during the past month of December, but there is still little certainty about how deadly or contagious it is. Despite everything, the worrying numbers – to date, more than 40,000 people have been infected and around 900 have died.
John Brownstein, head of the innovation department at Harvard Medical School, is part of the international team that daily searches for signs that might indicate the presence of the virus in regions where authorities have yet to detect it. Brownstein, who is also an expert in collecting data on social networks to identify trends in the health sector, now dedicates part of his day to analyzing posts, news and data that go beyond public agencies to draw conclusions about the virus path.
Currently, the team searches for references to specific symptoms, such as respiratory problems and fever, and segments the search based on the regions where suspected cases have been reported. The technology used can distinguish when a person claims to feel something when a user is just commenting on the topic – thus, false positives are excluded.
Speaking to Wired, Brownstein explains that these efforts are essential for authorities to be able to conclude knowingly how to allocate the resources available to manage the epidemic.