SYSTEM AND METHOD FOR TRAINING MACHINE LEARNING MODELS FOR RANKING SEARCH RESULTS Russian patent published in 2024 - IPC G06F16/9538 G06F16/901 G06F18/214 G06N20/00 

FIELD: data processing.

SUBSTANCE: invention relates to methods and systems for training and using machine learning models for ranking search results. Method includes: obtaining first history data for past searches performed by the user during the past period of time; obtaining second history data, part of which contains data of past searches performed by the user during the past user session; joint training of the first and second machine learning models to rank digital documents of the use stage, which includes training, based on the first history data of the first machine learning model, to form a vector representation of the first history data and training, based on the vector representation of the first history data and the second history data, of the second machine learning model to determine the value of the probability of the user’s action with the digital document of the use stage.

EFFECT: high accuracy of ranking a set of digital documents with limited server resources.

18 cl, 7 dwg