Israeli team says it has developed software to spot depressed bloggers

Inventors say program could enable mental health workers to identify individuals in need of treatment and recommend they seek help.

(HAARETZ)   Israeli researchers have developed software that claims to identify depressed bloggers by analyzing their writing.

The program scours blogs for words and phrases, descriptions and metaphors that can indicate the writer’s psychological state.

The software’s initial test run, which was part of a research study headed by Professor Yair Neuman of Ben-Gurion University’s department of education, combed more than 1,000 blog posts written by American bloggers that were online in 2004.

As part of the research, the software was asked to determine what it perceived as the 100 “most depressed” bloggers and the 100 “least depressed.”

Neuman told Haaretz that the software diagnoses largely matched those of four clinical psychologists who made their own diagnoses based on the blog posts.

“We found an 80 percent match between the automatic identification mechanism of the software and the human diagnosis given by the psychologists,” Neuman said.

“A psychologist knows how to spot various emotional states through intuition,” he said. “Here we have a program that does this methodologically through the innovative use of ‘web intelligence.’”

Neuman said the software could enable mental health workers to identify individuals in need of treatment and to recommend that they seek help.

“What does all of this mean from a practical standpoint?” he asked. “First of all, it shows that the technology is here and available and that it could be put to use.”

“In the United States there is a wide-ranging problem with depression,” said Neuman. “Through this software it will be possible to contact a blogger and request a general examination of the contents of his blog. If the blogger agrees, he will know whether he needs to seek professional counseling for any possible distress.”

Neuman said the researchers had received permission to analyze the blog posts.

The research and development for the software was funded by the Defense Ministry, yet Ben-Gurion University officials said yesterday the project would not be used for military purposes.

The program is capable of spotting words that express various emotions, like the names of colors that the writer employs to metaphorically describe certain situations. Hence words like “black,” if combined with other terms that describe such symptoms of depression as sleep deprivation and loneliness will be recognized by the software as “depressive” texts.

The software can also spot love and vengefulness (or at least thinks it can ).

Men who write prose laden with imagery from nature as well as words like “fire” or “lightning” could be determined by the program to be in love, as could women citing poetry or words related to music.

“The software does not rely on a single context-dependent word, but on a series of words strung together, terms and images chosen by the writer,” said Neuman.

…

Neuman cautioned against utilizing the technology for corporate purposes.

“I will not be pleased if this is put to negative use, like advertising for a certain product,” he said. “But I am all for using it as a means to spot cases of emotional distress.”