Randy Atkins: Google already uses web search data to estimate flu activity…but Mark Dredze, a computer scientist at Johns Hopkins, says social media – like Twitter broadcasts complaining about illness – may be more direct.

Mark Dredze: We think, in general, this is a better way of getting more detailed information about people because they’re sharing information instead of searching for information.

Randy Atkins: While humans can easily tell the difference between tweets like “I have the flu” and “I hope I don’t get the flu,” Dredze is training computers to do that on a massive scale.

Mark Dredze: To automate that process we rely on natural language processing which takes text that people write and we’re analyzing it for what the intention of the statement is.

Randy Atkins: Then geo-locate and tally people’s comments about illnesses, and Dredze says you should know where outbreaks are occurring. With the National Academy of Engineering, Randy Atkins, WTOP News.

• Hopkins Social Media and Health Research Group