Business Today - - THE BREAKOUT ZONE -

PEO­PLE SUF­FER­ING from de­pres­sion do not al­ways ad­mit to it – nei­ther to doc­tors nor fam­ily and friends and at times, not even to them­selves. It is of­ten typ­i­cal of the med­i­cal con­di­tion. Some­times, de­pres­sion is dif­fi­cult to de­tect, es­pe­cially in chil­dren and teenagers. So, re­searchers at MIT Me­dia Lab thought a bit of help from ar­ti­fi­cial in­tel­li­gence could be use­ful in such cases. Ac­cord­ingly, a re­search team is try­ing to teach an al­go­rithm to pick up cues from the speech of a group of peo­ple, some of whom are suf­fer­ing from de­pres­sion. Dif­fer­ences in tone and in­flex­ion and what is be­ing said are de­tected by the al­go­rithm, which (post-train­ing) has man­aged to pro­vide a de­pres­sion di­ag­no­sis with 77 per cent ac­cu­racy. That is not above-av­er­age, and the AI sys­tem must work on it to reach the point where its con­tri­bu­tions will be re­ally use­ful. Mean­while, there will be dan­gers of false pos­i­tives and misses be­cause de­pres­sion and hu­man be­hav­iour are in­cred­i­bly com­plex, unique and in­volve a mul­ti­tude of vari­ables.

Newspapers in English

Newspapers from India

© PressReader. All rights reserved.