Putting bloggers on the couch

BGU algorithm detects depression from online text.

By
June 17, 2010 14:07
2 minute read.
Putting bloggers on the couch

woman depression 88. (photo credit: )

It won’t yet replace a living and breathing psychologist or psychiatrist, but an algorithm developed at Ben-Gurion University’s education department claims to be able to detect depression with high accuracy from text written by people online.

Prof. Yair Neuman, who in fact has a degree in psychology, will present his BGU team’s work on the subject at the 2010 IEEE/WIC/ACM International Conference on Web Intelligence and Intelligent Agency Technology in Toronto in August.

Be the first to know - Join our Facebook page.


Working for nearly a year, the multidisciplinary team harvested Web intelligence and studied negative feelings expressed verbally and writing in English by some 300,000 Internet users. Using an English-only database of words, the team obtained from computer and then human analyses (for "fine tuning") the significance of simple words such as "revenge" or "depression."

The system improved by 10 percent the diagnoses of depression based solely on text. In addition, when checked on a huge collection of blogs, it identified the 1,000 most-depressed posts. Of these, it identified the 100 most-depressed and 100 least-depressed texts. When four clinical psychologists were given the written material to diagnose depression among the writers, the computerized system’s judgment was 80% accurate in comparison to the humanly judged texts.

The study was financed with an academic grant from the Defense Ministry, even though the work is relevant only to English texts so far. "We don’t yet have tools for the analysis of words in other languages," Neuman told The Jerusalem Post.

He noted that when people describe revenge, for example, many of them use positive words with cathartic imagery. The word surprisingly evoked more positive words than "love."

Another Israeli study has been grounded on similar databases of words; one recent one at the Hebrew University claims to identify sarcasm from possible ironic messages on the Internet. all of text from web.



"We are in the education department, but education is not just teaching in schools," Neuman said. "We search for hidden dimensions of meaning. It is also human development, the way human beings construct meaning."

Although conducted for academic purposes, the findings could be used to screen for would-be suicides -- if the text were written in English. Israel’s ERAN (Emotional First Aid) operates over the phone and not online. It could also be used for online counselling of the emotionally troubled and a variety of other applications. He will not only deliver a lecture at the Canada conference but also have his paper published in the conference’s proceedings.


Related Content

[illustrative photo]
September 24, 2011
Diabetes may significantly increase risk of dementia

By UNIVERSITY OF MICHIGAN HEALTH SYSTEM