Method, computing device and storage medium for post recall

By generating directed graphs and neural network models, using the similarity calculation of user and post embedding representation vectors, the problem of insufficient number of high-correlation posts in traditional post recall methods is solved, and richer association relationship learning and efficient recall are achieved.

CN114881114BActive Publication Date: 2024-11-05SHANGHAI SOULGATE TECH CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202210349370.0
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2022-04-01
Publication Date
2024-11-05
Estimated Expiration
2042-04-01

AI Technical Summary

Technical Problem

Traditional post recall methods are difficult to increase the number of posts with high correlation, user behavior data is sparse and collaborative filtering algorithms lack learning ability, resulting in a small number of recalled posts and a low correlation.

Method used

By obtaining the user's first interactive behavior data and the second interactive behavior data, a directed graph is generated, a random walk collects node sequences are used to generate input feature sequences, and a neural network model is used to generate user embedding representation vectors and post embedding representation vectors. Recall posts based on similarity calculation results.

Benefits of technology

It effectively increases the number of recalled high-relevant posts, learns more levels of relationship between users and posts, obtains richer relationships, and adjusts the node relationship level of the training sample.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN114881114B_ABST
    Figure CN114881114B_ABST
Patent Text Reader

Abstract

Embodiments of the present disclosure relate to a method, a computing device, and a computer storage medium for post recall. The method includes: obtaining first interactive behavior data of users for posts within a predetermined time interval, as well as second interactive behavior data between users, user identifiers, and post identifiers; generating a directed graph, the directed graph indicating multiple user nodes, multiple post nodes, directed edges for indicating first interactive behaviors, and directed edges for indicating second interactive behaviors; randomly walking on the directed graph in an alternating manner to collect multiple node sequences for generating input feature sequences; sliding the input feature sequence with a window of a predetermined size to generate multiple subsequences for training a neural network model; generating user embedding representation vectors and post embedding representation vectors via the trained neural network model to recall posts about the current user. The present disclosure can effectively increase the number of highly relevant posts recalled.
Need to check novelty before this filing date? Find Prior Art