Researchers from Stanford and Cornell universities have developed an algorithm that allows to detect Internet"trolls" before their activity will become apparent to other users. The system has high efficiency, making mistakes once out of five.

Read more...

Researchers from Stanford and Cornell universities have developed an algorithm that allows to detect Internet"trolls" before their activity will become apparent to other users. The system has high efficiency, making mistakes once out of five.

For algorithm development scientists analyzed the reports of several popular forums and news resources that have the ability to comment on published material. In the field of view of researchers got 1.7 million users, 40 million comments and 100 million "likes" and "dislikes" of communications, writes Quartz.

to identify the patterns, scientists have found what users were subsequently "banned" for their activity and to whom no sanctions were applied. Experts note that to define "Troll" is becoming increasingly difficult due to the changes in the style of communication of Internet users in General.