Information display method and device

Abstract

The invention discloses an information display method and a device, and the method comprises the steps: determining a sample user feature representing the preference of a user for the display form ofinformation according to the attribute of the user and the display form of each piece of sample information himself / herself clicked by the user; and taking the display form characteristics of the display form when the to-be-displayed information of each sample which is clicking by the user in history as a positive sample, and taking the display form characteristics of the display form when the to-be-displayed information of the non-clicked sample of the information corresponding to the clicked display form is not clicked as a negative sample. And jointly training a first model used for determining the user characteristics and a second model used for determining the display form characteristics through the determined sample user characteristics, positive samples and negative samples. Furthermore, the display form of the information can be determined by comparing the similarity between the user characteristics output by the first model and the display form characteristics output by the second model. According to the method in the specification, the CTR prediction model does not need to be operated on line, and the information display efficiency is effectively improved.

Description

technical field [0001] This specification relates to the Internet field, and in particular to an information display method and device. Background technique [0002] At present, in some e-commerce platforms, such as conversational ordering platforms, in order to guide users to consume, they usually display the corresponding information of merchants and products to users, help users find the merchants or products they are looking for, and finally complete the order. [0003] In the prior art, the e-commerce platform recalls various information corresponding to the information display request on the server side in an online manner according to the user's information display request, and each information can correspond to several display forms. Then predict the click through rate (CTR) that can be obtained by each display form online. Then, according to the predicted result, the display form that is most likely to be clicked by the user is screened out, and the information is ...

AI Technical Summary

Problems solved by technology

[0004] However, in actual usage scenarios, there are a large number of display forms of information recalled in response to user information requests, and online calculations are performed for each display form corresponding to the information, which will cause the e-commerce platform to display information to users. During the process, the amount of data processed online is large, the load of online processed data is large, and the model structure for predicting CTR is relatively complex, which affects the efficiency of information display

Images