The Odnoklassniki social network has launched a model based on machine learning that limits unauthorized dating and communication with unwanted interlocutors on the social network.
OK users who often refuse communication and dating will no longer be able to receive information about the activities of accounts that force online communication. The number of conversations cyber harassers have with users they do not know will be limited.
To protect users, OK analyzes social interaction between unfamiliar users (friend and friend requests, page visits, gifts and many other parameters) in real time and categorizes them. The neural network regularly finds accounts that add people they don’t know in common as friends, send gifts to strangers in bulk, tutor them, and write personal messages. The OK pattern often involves people rejecting friend requests, blacklisting users, rejecting gifts, or complaining about strangers’ actions.
Based on the collected data, OK limits interaction between these groups of accounts and hides notifications from unauthorized interlocutors, for example, about friend requests, classes, gifts, page visits in photos or posts. The model also reduces the number of possible conversations between these user groups; The number of conversations that those who send private messages in bulk will be limited.
The model was tested on different audiences over several months, and as part of the experiment, the number of complaints about intrusive attention from unfamiliar accounts decreased by 25% compared to previous periods.
The introduction of new machine learning technology will make interaction within the social network more comfortable and help maintain a friendly atmosphere in OK.