Topic-aware heterogeneous graph neural network for link prediction

• Main Authors: XU, Siyong, YANG, Cheng, FANG, Yuan, TIANCHI, Yang, ZHANG, Luhao
• Format: text
• Language: English
• Published: Institutional Knowledge at Singapore Management University 2021
• Subjects: Heterogeneous graph; Graph neural networks; Representation learning; Link prediction; Databases and Information Systems
• Online Access: https://ink.library.smu.edu.sg/sis_research/6805; https://ink.library.smu.edu.sg/context/sis_research/article/7808/viewcontent/123.pdf
• Institution: Singapore Management University

Heterogeneous graphs (HGs), consisting of multiple types of nodes and links, can characterize a variety of real-world complex systems. Recently, heterogeneous graph neural networks (HGNNs), as a powerful graph embedding method to aggregate heterogeneous structure and attribute information, has earned a lot of attention. Despite the ability of HGNNs in capturing rich semantics which reveal different aspects of nodes, they still stay at a coarse-grained level which simply exploits structural characteristics. In fact, rich unstructured text content of nodes also carries latent but more fine-grained semantics arising from multi-facet topic-aware factors, which fundamentally manifest why nodes of different types would connect and form a specific heterogeneous structure. However, little effort has been devoted to factorizing them.In this paper, we propose a Topic-aware Heterogeneous Graph Neural Network, named THGNN, to hierarchically mine topic-aware semantics for learning multi-facet node representations for link prediction in HGs. Specifically, our model mainly applies an alternating two-step aggregation mechanism including intra-metapath decomposition and inter-metapath mergence, which can distinctively aggregate rich heterogeneous information according to the inferential topic-aware factors and preserve hierarchical semantics. Furthermore, a topic prior guidance module is also designed to keep the quality of multi-facet topic-aware embeddings relying on the global knowledge from unstructured text content in HGs. It helps to simultaneously improve both performance and interpretability. Experimental results on three real-world HGs demonstrate that our proposed model can effectively outperform the state-of-the-art methods in the link prediction task, and show the potential interpretability of learnt multi-facet topic-aware representations.