Global Context Enhanced Social Recommendation with Hierarchical Graph Neural Networks

Huance Xu,Chao Huang,Yong Xu,Lianghao Xia,Hao Xing,Dawei Yin
DOI: https://doi.org/10.48550/arXiv.2110.04039
2021-10-08
Abstract:Social recommendation which aims to leverage social connections among users to enhance the recommendation performance. With the revival of deep learning techniques, many efforts have been devoted to developing various neural network-based social recommender systems, such as attention mechanisms and graph-based message passing frameworks. However, two important challenges have not been well addressed yet: (i) Most of existing social recommendation models fail to fully explore the multi-type user-item interactive behavior as well as the underlying cross-relational inter-dependencies. (ii) While the learned social state vector is able to model pair-wise user dependencies, it still has limited representation capacity in capturing the global social context across users. To tackle these limitations, we propose a new Social Recommendation framework with Hierarchical Graph Neural Networks (SR-HGNN). In particular, we first design a relation-aware reconstructed graph neural network to inject the cross-type collaborative semantics into the recommendation framework. In addition, we further augment SR-HGNN with a social relation encoder based on the mutual information learning paradigm between low-level user embeddings and high-level global representation, which endows SR-HGNN with the capability of capturing the global social contextual signals. Empirical results on three public benchmarks demonstrate that SR-HGNN significantly outperforms state-of-the-art recommendation methods. Source codes are available at: <a class="link-external link-https" href="https://github.com/xhcdream/SR-HGNN" rel="external noopener nofollow">this https URL</a>.
Information Retrieval,Artificial Intelligence,Machine Learning
What problem does this paper attempt to address?
### The Problem Addressed by the Paper This paper aims to address two significant challenges in social recommendation systems: 1. **Multi-type User-Item Interaction Behaviors and Their Cross-Relation Dependencies**: - Most existing social recommendation models fail to fully explore the multi-type user-item interaction behaviors and their underlying cross-relation dependencies. For example, various rating behaviors of users in online review systems or different activities such as browsing and purchasing on e-commerce websites. - These multi-type relationships can provide auxiliary behavioral semantics, but existing social recommendation models find it difficult to comprehensively capture this information. 2. **Representation Capability of Global Social Context**: - Although the learned social state vectors can model pairwise dependencies between users, they are still limited in capturing the global social context among users. - The currently designed user social information embedding functions lack the ability to effectively encode high-order relational structures, which implies social similarity in user-user relationships. To address these challenges, the authors propose a novel social recommendation framework based on hierarchical graph neural networks (SR-HGNN). This framework improves existing social recommendation systems in the following ways: - **Relation-aware Reconstructed Graph Neural Network**: Injects cross-type collaborative semantics and automatically extracts multi-relation collaborative signals from user-item interactions. - **Global Social Context Enhanced Social Relation Encoder**: Enhances the model's ability to capture global social context signals based on a mutual information learning paradigm between lower-level user embeddings and higher-level global representations. Through these methods, SR-HGNN has been experimentally proven to significantly outperform existing recommendation methods on three public benchmark datasets.