With 230 million active users, Twitter is a great for observing human behavior and perhaps getting to the root of public health problems. Researchers at Microsoft Research Redmond wondered if they could scan tweets in order to detect if someone is depressed.
Microsoft Research Redmond Co-Director Eric Horvitz and a team of researchers developed a model that scans through tweets and predicts depression in the user. They claim, while it is not quite perfect, the accuracy is close to 70 percent.
What Were The Researchers Looking For When Scanning Twitter For Depressed Users?
Horvitz and his team signed up 637 Twitter users to be part of the study. The researchers went back into users’ histories as far as a year, looking for signs in the number of tweets made per day, the language used, when they were tweeting, the level of engagement and a variety of other factors.
What Did They Find?
The team noticed that depressed users were more likely to tweet during the hours of 9 p.m. and 6 a.m. These users also used a number of words that are indicative of depression such as anxiety, fatigue, addictive and attacks. Some of the words that were far more surprising consisted of favorite, wants, tolerance, love, songs and movies.
Users who were depressed tended to tweet less toward other people, and less overall. Granted, it is necessary to take into consideration a users’ entire history as there might be an individual who tweets about anxiety and love at 3 a.m., but that doesn’t mean they are depressed.
How Findings May Be The Beginning Of New Ways Of Analyzing People’s Feelings As Well As Preventing Suicide
Horvitz thinks this type of research will become invaluable when measuring reactions to major events. Until recently there hasn’t been a single type of analysis at this level. It could open up a whole new way of viewing how individuals are affected by the news during and after intense events.
The research team is attempting to improve on this model by correcting for false and misleading data, and extending the list of trigger words in order to provide a real-time way of monitoring depression in the general population.
It will probably be a long time before this model is developed to the point where it can provide help to those suffering from depression. Horvitz is driven by the knowledge that 30,000 suicides due to depression happen every year. He hopes that one day his research will save at least a few of those individuals.
Discover How To Recognize Technology Addiction In Teens – Is Your Teen Addicted To The Internet?