Hopkins mines Twitter for real flu cases in real time
As flu infiltrates our homes and workplaces, researchers have developed a computer-based method of tracking Twitter posts that could alert the public to sickness faster than official health reports with greater accuracy than existing online tools. If mining the social networking site helps keeps people healthy, it'll be nothing to sneeze at.
The web has been awash with flu-tracking apps such as Google Flu Trends and other tools, yet the researchers from Johns Hopkins University believe that they have improved their ability to sort idle Twitter noise about the virus from actual cases. To prove it, they compared their results from November and December with official reports during the same period from the Centers for Disease Control and Prevention.
The researchers say their way of mining Twitter, using methods based on human language processing tech, bested accuracy from previous systems. What is more, Hopkins researchers say their method yields results in real time, much faster than the two weeks or so that the CDC takes to publish flu trends based on records from healthcare providers.
And Kobe doesn't count. At least when you or I tweet about someone else's illness, as was the case when Twitter was abuzz about Los Angeles Lakers star Kobe Bryant's bug, the Hopkins method picks up on the difference between, say, "Flu kept Kobe out of tonight's game, dude," and "Dude, I've got the flu and Kobe's no Jordan."
"Mr. Bryant's health notwithstanding," Hopkins researcher David Broniatowski stated, "such tweets do very little to help public health officials prepare our nation for the next big outbreak."
Even with the increasing accuracy of online trackers, however, the CDC sticks with its recommendation that people get flu shots. Meantime, the team from Hopkins is interested in applying their system to other diseases.
Special Report: Top 50 Industry Voices in Biotech Twittersphere -- 2012