In recent years, with the booming development of deep learning, it makes the judicial big data analysis technology increasingly attracts people’s attention, and gives rise to many new applications oriented to the judicial field, and the intelligent court project is also included in the judicial construction process. The research realizes the automatic classification and retrieval model of legal documents through the graph neural network-based algorithm, and the article proposes the DASA-GNN text classification model based on the Texting model, and adopts the EDA data enhancement method combined with the self-attention mechanism in the data sampling stage. Meanwhile, a graph convolutional network is introduced on the basis of the pre-trained model to learn the contextual information and global structured information of the text, and a text retrieval model based on the topological feature representation of the convolutional graph is proposed. Then the application effect of the classification model as well as the retrieval model is analyzed through experiments, and the results show that the Acc value of the DASA-GNN text classification model reaches 95.12% on the PKULawData dataset, which is improved compared with that of the baseline models such as BERT, DADGNN, and RCNN, etc.; finally, a comparison experiment is carried out on the legal retrieval dataset LeCaRD, and the experimental The results show that the text retrieval model based on the topological feature representation of convolutional graph has a better retrieval effect compared to other retrieval methods.
