logo


Engadget


Researchers are tracking your flu-infected tweets

Sick

First Posted: 01/31/13 04:36 PM ET Updated: 02/01/13 04:50 PM ET

By Rob Zanicchi posted Jan 31, 2013


We know that technology can play a major role in the fight against disease, but now it seems Twitter can help alleviate the pain as well. Meet Dr. Mark Dredze, a computer science professor at Johns Hopkins University and a research scientist at its Human Language Technology Center of Excellence -- he's led a small team in developing a tracking system that can utilize tweets to accurately map flu outbreaks and their severity across the US.

Beginning in 2009, Dredze and his fellow researchers wanted to see if Twitter could be utilized to obtain reliable health data, training a computer system to track related tweets about the topic. Detailed in a 2011 report, the group had sampled two billion tweets and the algorithm filtered out 1.5 million about health. Impressively, it's able to analyze 5,000 public tweets per minute, while learning human language patterns to help distinguish desirable data from noise -- as they put it, the difference between "Bieber-fever and actual fevers." As of late, they've refined the algorithm to be better than others by giving it the ability distinguish between people talking about a specific sickness versus folks who are actually infected with one. Going a step further, it can also breakdown data from anyone describing how sick they are. Interestingly, the twitter-sourced stats from November and December 2012 of the current flu season essentially echoed that of the Centers for Disease Control. It's a notable a feat, considering the results are real-time -- as the college notes, the hospital-sourced info used by the CDC takes about two weeks to compile.

Dredze and team decided to release the details about their filtering improvements due to the monstrous size of the current influenza outbreak, mapping it out against previous years. All in all, they hope to share the setup with top health agencies and eventually branch out to following other diseases. It seems likely, too, as the college mentions a contest to make similar software was put up last year by U.S. Department of Health and Human Services. Check out the video below for more, and seek comfort in knowing your 140-character complaints are now helping science.


Via: Mashable, Computerworld
Source: Johns Hopkins University