Content recommendation method and apparatus, device, medium, and program product

Patent metadata

Field	Value
Publication number	US-12579205-B2
Application number	US-202418939506-A
Country	US
Kind code	B2
Filing date	Nov 6, 2024
Priority date	Nov 29, 2022
Publication date	Mar 17, 2026
Grant date	Mar 17, 2026

Abstract

A content recommendation method is provided. A first feature representation of a first user and a second feature representation of a second user are obtained. A cluster center corresponding to the first user is determined. An extra-domain feature representation and intra-domain feature representation of the first user are obtained. The intra-domain feature representation is based on mapping the extra-domain feature representation with the mapping relationship function. A target feature representation of the first user is determined based on the intra-domain feature representation and the first feature representation. Target content that matches the target feature representation of the first user is determined. The target content is pushed to the first user.

First claim

What is claimed is:

1 . A content recommendation method, the method comprising:
obtaining a first feature representation of a first user in a first function platform, wherein the first feature representation of the first user is extracted from attribute data of the first user; obtaining a second feature representation of a second user in the first function platform, wherein the second feature representation of the second user is extracted from attribute data of the second user; determining a cluster center corresponding to the first user based on the first feature representation and the second feature representation; obtaining an extra-domain feature representation of the first user in a second function platform, wherein the extra-domain feature representation of the first user is extracted from first historical interaction data of the first user in the second function platform; obtaining a mapping relationship function corresponding to the cluster center, wherein the mapping relationship function indicates a feature representation relationship between the second function platform and the first function platform; determining an intra-domain feature representation of the first user, wherein the intra-domain feature representation is based on mapping the extra-domain feature representation with the mapping relationship function; determining a target feature representation of the first user based on the intra-domain feature representation and the first feature representation; determining, by processing circuitry and from a candidate content recommendation pool, target content that matches the target feature representation of the first user; and pushing the target content to the first user.

2 . The method according to claim 1 , wherein the obtaining the mapping relationship function further comprises:
performing parameter replacement on a pre-generated parameter-containing mapping function to obtain the mapping relationship function corresponding to the cluster center.

3 . The method according to claim 2 , wherein the performing the parameter replacement further comprises:
obtaining the pre-generated parameter-containing mapping function, the pre-generated parameter-containing mapping function including a to-be-filled parameter position; and replacing the cluster center with the to-be-filled parameter position to obtain the mapping relationship function.

4 . The method according to claim 3 , the method further comprising:
obtaining a sample intra-domain feature representation of a sample user, the sample user corresponding to a sample cluster center, wherein the sample intra-domain feature representation of the sample user is extracted from second historical interaction data of the sample user in the first function platform; obtaining third historical interaction data of the sample user in the second function platform, and extracting a feature based on the third historical interaction data, to obtain a sample extra-domain feature representation of the sample user; obtaining a sample intra-domain mapping feature corresponding to the sample user, wherein the intra-domain mapping feature is based on mapping the sample extra-domain feature representation of the sample user with a candidate mapping function; obtaining a reconstruction loss based on the sample intra-domain feature representation and the sample intra-domain mapping feature of the sample user; and training the candidate mapping function based on the reconstruction loss.

5 . The method according to claim 1 , further comprising:
performing clustering analysis on the second user based on the second feature representation, to obtain a plurality of clusters with a plurality of cluster centers, each cluster of the plurality of clusters including one cluster center of the plurality of cluster centers; and determining, based on similarity degrees between the first feature representation and the plurality of clusters, the cluster center of the plurality of cluster centers corresponding to the first user.

6 . The method according to claim 5 , wherein the performing the clustering analysis further comprises:
obtaining clustering information, the clustering information indicates position information of an initial cluster center; obtaining a similarity degree between the second feature representation and the initial cluster center; determining a first cluster distribution result based on the similarity degree between the second feature representation and the initial cluster center, the first cluster distribution result including a feature distribution corresponding to each initial cluster center; and performing discrete analysis on the first cluster distribution result, to obtain a second cluster distribution result; and determining the plurality of clusters based on the second cluster distribution result, the second cluster distribution result indicating each second feature representation with a corresponding cluster center.

7 . The method according to claim 5 , wherein the determining the cluster center of the plurality of cluster centers further comprises:
calculating a distance between the first feature representation and each cluster center of the plurality of clusters; and determining the cluster center of the plurality of cluster centers having a shortest distance as the cluster center corresponding to the first user.

8 . The method according to claim 5 , wherein the determining the cluster center of the plurality of cluster centers further comprises:
calculating the similarity degree between the first feature representation and each cluster center of the plurality of clusters; determining, based on the similarity degrees, probabilities that the first user belongs to each cluster center; and determining, based on a maximum probability from the determined probabilities, the cluster center of the plurality of cluster centers having a shortest distance as the cluster center corresponding to the first user.

9 . The method according to claim 1 , wherein the obtaining the extra-domain feature representation further comprises:
determining, based on the first historical interaction data, a target heterogeneous graph including a plurality of meta-paths, and the target heterogeneous graph indicating a historical interaction relationship between the first user and a platform element in the second function platform; and obtaining the extra-domain feature representation of the first user based on path feature representations corresponding to the plurality of meta-paths.

10 . The method according to claim 9 , wherein the determining the target heterogeneous graph further comprises:
obtaining node attention of a path node in the meta-paths, the path node indicating the platform element having the historical interaction relationship with the first user; and performing aggregation processing on the node attention, to obtain the path feature representations of the meta-paths.

11 . A content recommendation apparatus, comprising:
processing circuitry configured to:
obtain a first feature representation of a first user in a first function platform, wherein the first feature representation of the first user is extracted from attribute data of the first user;
obtain a second feature representation of a second user in the first function platform, wherein the second feature representation of the second user is extracted from attribute data of the second user;
determine a cluster center corresponding to the first user based on the first feature representation and the second feature representation;
obtain an extra-domain feature representation of the first user in a second function platform, wherein the

Assignees

• Tencent Tech Shenzhen Co Ltd

Inventors

• Zheng Jiawei
• Gu Hao
• Yi Lingling

Classifications

• Y02D10/00

  Energy efficient computing, e.g. low power processors, power management or thermal management · CPC title

• G06F18/23213

  with fixed number of clusters, e.g. K-means clustering · CPC title

• G06F16/958

  Organisation or management of web site content, e.g. publishing, maintaining pages or automatic linking · CPC title

• G06F16/9535 Primary

  Search customisation based on user profiles and personalisation · CPC title