Chinese Text Classification Method Based on Double Pre-Processed Graph Aggregation Network

To address the problem of insufficient node feature recognition and shallow network layers in Chinese text classification, this paper proposes a Chinese text classification method called Dual Preprocessing Graph Information Aggregation Network (DPGIAN). This method solves the problem of insufficient initial feature recognition of nodes and the inability of edge weights to highlight the weight ratio between keywords through a dual preprocessing strategy of node preprocessing and edge weight preprocessing; At the same time, a deeper graph information aggregation feature extraction method was used to solve the problem of excessive feature smoothing when obtaining deep semantic features in graph structured networks, improving the accuracy of Chinese text classification. By conducting experiments on multiple Chinese text classification datasets and comparing the experimental results with multiple mainstream text classification methods on the same dataset, our method outperforms the current mainstream text classification methods on the same dataset. Multiple ablation comparison experiments were also conducted to further verify the effectiveness of our method.