Web Crawlers for News Aggregation Systems based on Graph Convolutional Neural Networks

The author delves into the news aggregation system web crawler KMGCN based on graph convolutional neural networks. KMGCN effectively improves the efficiency and accuracy of news aggregation by constructing multimodal content graphs and utilizing knowledge driven graph convolutional networks. This study evaluated the performance of KMGCN on different datasets and validated its superiority by comparing it with benchmark methods. This study explores the performance of a news aggregation system web crawler based on graph convolutional neural networks when the feature dimension changes. By adjusting the feature dimension d, it was observed that the accuracy of KMGCN slightly improved when d increased from 50 to 250 on both datasets. However, as d further increased to 300, the accuracy on one of the datasets slightly decreased. For the F1 indicator, increasing d to 200 did not bring any improvement. After comprehensive analysis, determine that the word vector dimension d in the text is 100 to optimize the performance of KMGCN. Overall, the experimental results of this study demonstrate the effectiveness of the method on two common datasets. KMGCN has significant advantages in handling multimodal content, filtering noise, and outliers, providing new ideas and methods for the development of news aggregation.