Sample data set acquisition method and apparatus, equipment and storage medium

Abstract

The invention discloses a sample data set acquisition method and apparatus, equipment and a storage medium, belonging to the technical field of the Internet. The method comprises the steps of acquiring a first sample data set corresponding to any search term, and selecting at least one piece of target negative sample data according to the position of initial positive sample data in a same search result interface to which the initial positive sample data belong in the same search result interface; and according to the historical click rate of a user identifier, selecting at least one piece of target positive sample data, and forming a second sample data set corresponding to the any search term by using the target negative sample data and the target positive sample data, and applying the second sample data set to training of a sorting model. The number of the negative sample data is reduced, the situation that the number of the negative sample data is far larger than the number of the positive sample data is avoided, and therefore, the situation that the sorting model of follow-up training is closer to the characteristics of the negative sample data is avoided. The second sample dataset is subsequently adopted to train the sorting model, so the accuracy of the sorting model is improved.

Description

technical field [0001] The present application relates to the technical field of the Internet, and in particular to a method, device, equipment and storage medium for acquiring a sample data set. Background technique [0002] In order to ensure the accuracy of the search results, the sorting model is usually invoked during the search to sort the multiple pieces of data obtained from the search. How to train an accurate ranking model has become an urgent problem to be solved. [0003] In related technologies, for any user, when the user searches based on the search term and obtains at least one piece of data, each piece of data displayed can be regarded as the data seen by the user and can be used as sample data. If there is one piece of data, record the data as positive sample data. If the user does not click on any of the displayed data, record the data as negative sample data. Using the above method, positive sample data and negative sample data can be obtained based on t...

AI Technical Summary

Problems solved by technology

[0004] However, since the number of data browsed by users will be much greater than the number of clicked data, the positive sample data obtained by the above method is far less than the negative sample data. In the process of training the ranking model, the features learned by the ranking model will be more biased towards negative sample data. The characteristics of the sample data lead to low accuracy of the trained ranking model

Images