Abstract:Graph convolutional networks (GCNs), aiming to obtain node embeddings by integrating high-order neighborhood information through stacked graph convolution layers, have demonstrated great power in many network analysis tasks such as node classification and link prediction. However, a fundamental weakness of GCNs, that is, topological limitations, including over-smoothing and local homophily of topology, limits their ability to represent networks. Existing studies for solving these topological limitations typically focus only on the convolution of features on network topology, which inevitably relies heavily on network structure. Moreover, most networks are text-rich, so it is important to integrate not only document-level information, but also the local text information which is particularly significant while often ignored by the existing methods. To solve these limitations, we propose BiTe-GCN, a novel GCN architecture modeling via bidirectional convolution of topology and features on text-rich networks. Specifically, we first transform the original text-rich network into an augmented bi-typed heterogeneous network, capturing both the global document-level information and the local text-sequence information from texts. We then introduce discriminative convolution mechanisms, which performs convolution on this augmented bi-typed network, realizing the convolutions of topology and features altogether in the same system, and learning different contributions of these two parts (i.e., network part and text part), automatically for the given learning objectives. Extensive experiments on text-rich networks demonstrate that our new architecture outperforms the state-of-the-arts by a breakout improvement. Moreover, this architecture can also be applied to several e-commerce search scenes such as JD searching, and experiments on JD dataset show the superiority of the proposed architecture over the related methods.

GCN-CA: Co-Attention for Text Matching by Text Graph Representation

Mixed Geometry Message and Trainable Convolutional Attention Network for Knowledge Graph Completion

Multiresolution Graph Attention Networks for Relevance Matching

Matching Long Text Documents Via Graph Convolutional Networks.

Feature Interaction Based Graph Convolutional Networks for Image-Text Retrieval

MFG-R: Chinese Text Matching with Multi-Information Fusion Graph Embedding and Residual Connections.

Cross Attention Graph Matching Network for Image-Text Retrieval

Cross-modal Graph Matching Network for Image-text Retrieval

BiTe-GCN: A New GCN Architecture Via Bidirectional Convolution of Topology and Features on Text-Rich Networks

Neural Graph Matching Networks for Chinese Short Text Matching.

Graph Structured Network for Image-Text Matching

Text-Image Matching for Cross-Modal Remote Sensing Image Retrieval via Graph Neural Network

Learning for Feature Matching Via Graph Context Attention

Novel GCN Model Using Dense Connection and Attention Mechanism for Text Classification

Understanding Graph Convolutional Networks for Text Classification

GLMNet: Graph Learning-Matching Networks for Feature Matching

Topic-aware Cosine Graph Convolutional Neural Network for Short Text Classification

CGNN: Caption-assisted Graph Neural Network for Image-Text Retrieval

Text Matching As Image Recognition

The Study on the Text Classification Based on Graph Convolutional Network and BiLSTM

Contextual Relation Guided Image-Text Matching