A news trend prediction system and method based on graph convolutional neural network

Through a news trend prediction system based on graph convolution neural network, the Chinese pre-trained language model and graph convolution neural network extract news text features are solved, and the problem of insufficient Chinese text prediction in the existing technology is achieved, and efficient and accurate news trend prediction is achieved.

CN114021550BActive Publication Date: 2025-03-25CHENGDU CAS INFORMATION TECH CO LTD
3 Cites 0 Cited by

Patent Information

Application Number
CN202111298567.8
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2021-11-04
Publication Date
2025-03-25
Estimated Expiration
2041-11-04

AI Technical Summary

Technical Problem

When using Chinese text to predict news trends, the prior art faces the problem that pre-trained language models are insufficient in processing ultra-long texts, lack of a database with complete labeling, and the inability of pure deep neural networks to meet the prediction needs.

Method used

A news trend prediction system based on graph convolution neural network is adopted, and feature extraction, database construction and trend matching are performed through news text input unit, news feature extraction unit, news trend database and news trend matching unit, and feature vectors of news text are extracted using Chinese pre-trained language model and graph convolution neural network.

Benefits of technology

Accurate prediction of news trends is achieved, the gap in the Chinese news trend database is filled, the prediction accuracy and speed is improved, and the news trend analysis of different application scenarios is supported through the construction of the feature vector matrix.

✦ Generated by Eureka AI based on patent content.
Patent Text Reader

Abstract

The present invention discloses a news trend prediction system and method based on a graph convolutional neural network, and relates to the field of neural networks. The present invention extracts features from prior news texts with development trend annotations and establishes a news trend database. It only needs to calculate the feature vector of the target news text and find the most similar news text in the database, so that the news development trend of the target news text can be predicted by the most similar news text; by establishing a news trend database with complete annotations, the gap in the current status and trend database of Chinese news is filled, and the news trend database can be connected to different application scenarios, thereby promoting the development of the field of news trend prediction; the embedded vector of the named entity is obtained by using the Chinese pre-trained language model as the node initial value of the graph convolutional neural network, and the graph convolutional neural network is used to extract the features of the entire graph, and the graph embedding vector can better retain the hierarchical features through the graph pooling algorithm.
Need to check novelty before this filing date? Find Prior Art