The system will continue to be improved
The team of the largest Russian online trading platform Wildberries shared news regarding the situation with fictitious negative reviews that were created by competitors of sellers and scammers. As the press service said, Wildberries tested machine learning of an artificial intelligence system to hide reviews from «fictitious buyers».
During 1.5 months of testing algorithms, Wildberries hid more than 150 thousand negative reviews left by «fictitious buyers». According to the developers, this made it possible to minimize the human factor, objectively check the content of comments and remove compromised ratings.
Wildberries said:
In November, we improved the automated search model and, in addition to reviews, began to analyze photos in comments for spam.
According to preliminary results, the negative reviews turned out to be the result of unfair competition. More than 400 thousand hidden photos turned out to be spam with fraudulent links. Such spammers make money by generating negative reviews and the gullibility of those who, on their behalf, leave dishonest comments.