Researchers say language cues in tweets are providing new insights into mental health disorders. (Photo by Justin Sullivan/Getty Images)

Language reveals a lot about a person's mental state. So researchers are hoping the messages people share on Twitter will shed new light on mental illness research.

Johns Hopkins University researchers say language clues provided through Twitter and other social media platforms can help provide a deeper understanding of mental illness and where it exists. By mining Twitter for language cues, they're hoping to fill the data gap that mental health researchers face in trying to diagnose and understand many disorders.

"So much of mental health and psychology is based on language and an analysis of language," said Glen Coppersmith, a senior research scientist and member of the Johns Hopkins research team that previously used social media to track another pressing public health concern — the spread of flu season.

The researchers analyzed public tweets between 2008 and 2013 from people who said they had been diagnosed with mental illnesses. Their analysis included people who referenced having bipolar disorder, major depressive disorder, post-traumatic stress disorder and seasonal affective disorder, while filtering out messages that didn't appear to be legitimate.

The researchers point out a few potential drawbacks with this approach. Given the existing social stigma around mental illness, those speaking about it on Twitter and other social media platforms may not be representative of the population. There's also no way to determine whether the diagnoses are accurate, and Twitter itself isn't representative sample. Still, they found some cues in the language that they say could help mental health professionals improve treatments of find new avenues for treatments.

"It looks like there's a whole lot of very subtle things that any one of them alone doesn't tell you too much, but when you look at them together, you're able to see some patterns emerging." Coppersmith said.

One earlier analysis from the Johns Hopkins team showed increased rates of PTSD at military bases with more frequent deployments to Iraq and Afghanistan. The finding itself isn't so surprising, since military members have higher higher rates of PTSD compared to the general public. But the finding, which the researchers said was the first social media analysis of individuals with PTSD, showed how public health officials could use Twitter to quickly understand how to direct resources following traumatic incidents.

Another early finding showed people who identify as having depression say "I" more than the rest of the population. It shows that they're more introspective, or they may be spending more time alone. It's hard to say what this signal alone would mean for mental health diagnosis and treatment going forward, Coppersmith said, but it could inform how mental health professionals think about linguistic signals.

"This could indicate perhaps different, better questions, updating questionnaires made in the '70s and '80s to assess these mental health problems," he said. "Maybe this will point to some ways we can get further insight into mental illness."

The Johns Hopkins researchers aren't alone in their efforts to decode psychological clues in social media messages. For example, the World Well-Being Project at the University of Pennsylvania last year linked language use to personality traits in a massive study of Facebook statuses of 75,000 people.