Information recommendation model training method and device, information recommendation method and device, and equipment

Abstract

The invention provides an information recommendation model training method and device, an information recommendation method and device, and equipment, and relates to the technical field of deep learning. The method comprises the following steps: acquiring a multi-level label tree of the application field of an information recommendation model, wherein each level of the multi-level label tree comprises at least one label feature of the application field, and in every two adjacent levels of the multi-level label tree, one label feature of the next level uniquely belongs to one label feature of the previous level; generating an embedded matrix according to the label feature of the last level in the multi-level label tree; generating an index record table corresponding to the embedded matrix according to the affiliation relation between the label features in the multi-level label tree, wherein the index record table comprises at least one index value; and training to obtain an informationrecommendation model according to the embedded matrix and the index record table. According to the scheme, it can be ensured that the information recommendation model quickly and accurately learns therelevance and consistency among the multi-level label features.

Description

technical field [0001] The present application relates to the field of deep learning technology, in particular, to an information recommendation model training method, information recommendation method, device and equipment. Background technique [0002] The rapid development of the Internet has brought a lot of convenience to people's life, but also put users in the dilemma of information overload. Therefore, a personalized recommendation algorithm can be used to recommend a large number of online resources to users who may be interested, so that users can quickly obtain valuable information from a large amount of data. [0003] At present, most of the personalized recommendation algorithms construct a hierarchical relationship for the extracted label features, and construct an embedding matrix for each level of label features, and use these embedding matrices to train a personalized recommendation model. [0004] However, using the existing method to construct multiple em...

AI Technical Summary

Problems solved by technology

[0004] However, using the existing method to construct multiple embedding matrices for different levels of label features is to consider the label features between levels separately, which makes the multiple embedding matrices constructed have the problem of fragmentation, which leads to the inability of the personalized recommendation model to learn Correlation between multi-level label features

Images