METHOD AND SERVER FOR REPEATED TRAINING OF MACHINE LEARNING ALGORITHM Russian patent published in 2021 - IPC G06N20/00 G06F16/906 G06F16/95 

FIELD: information technology.

SUBSTANCE: invention relates to method and server of repeated training of machine learning algorithm (MLA). Method comprises invoking a server through a communication network to a resource comprising a plurality of content items, each of which was sent to this resource by at least one user; identifying based on a first set of features a content item from a plurality of content items associated with a document, represented by a search engine server as a search result in response to a search query; analysis based on feature set of document of content element to determine whether this document was incorrectly classified by MLA algorithm in response to search request; and if it is determined that the document was incorrectly classified, forming a training object based on that document, including assigning a label to the document in form of an indication of incorrect classification, and repeated training of MLA algorithm on the basis of this training object.

EFFECT: technical result is higher efficiency of training MLA by performing repeated training.

28 cl, 7 dwg