VKontakte is testing a neural network to combat hate speech

The social network VKontakte is working on artificial intelligence to automatically recognize materials with hostile statements. “Such Affairs” was informed about this by the site’s press service.

The new neural network will help quickly find materials with hostile statements and publications that provoke the audience to show aggression. Also, with the help of this technology, VKontakte hopes to speed up the processing time for complaints and help users of the social network less often encounter manifestations of intolerance.

The neural network is taught to detect different types of insults and threats: manifestations of xenophobia, racism, homophobia, sexism, as well as hostility on religious grounds, due to age, appearance or the presence of diseases. In order to train artificial intelligence, VKontakte added a new category of complaints for users - “Hostile Statements.” You can select it when you click the “Report” button on the offensive publication.

“We share the position of our users and partners regarding hate speech and the glorification of violence. If you notice an offensive publication, report it. It is more effective to fight manifestations of intolerance together - and we urge users to join us in this fight,” said Ivan Korneev, head of the VKontakte moderation team.

In the first half of the year, the social network deleted 520 thousand publications due to incitement to hatred, and blocked 1,340 profiles and 2,470 communities for disseminating hostile statements.