Scientific Paper Classification by Fusing BERT and GCN
Xiaohe Zhang, Xinguo Yu, Xiaoqian Liu, Xiaopan Lyu · 2023
This paper proposes a BERT-based graph convolutional neural network (BERT-GCN) model for classifying scientific papers, which combines a fine-tuned BERT and a graph convolutional neural network to enhance the performance of classification. To enhance the generalization ability of BERT-GCN, this paper proposes using BERT as the basic model and combining three mechanisms - span masking, learning rate attenuation, and data augmentation - to fine-tune the BERT model for scientific paper classification. To ensure classification performance, this paper uses a graph convolutional neural network to capture the structural features of paper titles. Experimental results obtained from the test on two benchmark datasets validate the superiority of the proposed BERT-GCN model that achieves state-of-the-art performance algorithms against the baseline models.