Geometric-Contextual Mutual Infomax Path Aggregation for Relation Reasoning on Knowledge Graph

Xingrui Zhuo,Gongqing Wu,Zan Zhang,Xindong Wu

DOI: https://doi.org/10.1109/tkde.2024.3360258

IF: 9.235

2024-01-01

IEEE Transactions on Knowledge and Data Engineering

Abstract:Relation reasoning in Knowledge Graph Completion (KGC) aims at predicting missing relations between entities. Recently, effective KGC methods have usually focused on exploring the path pattern between entities, such as reward-based path walking and path context mining, to complete target relations. However, these methods typically suffer from two challenges: 1) They have difficulty in handling the individual representation limitation of candidate paths when there are no paths that directly represent latent relations between entities; 2) They overlook the biases of path context induction, which leads to unreasonable information interfering with the model's reasoning. To manage these challenges, a Geometric-Contextual Mutual Infomax (GCMI) path aggregator is proposed for relation reasoning. First, we design an attentive path aggregator with a shared Transformer encoder to capture the contexts from several candidate paths parallelly and integrate these contexts to sufficiently represent the latent relations of each entity pair for reasoning. Then, the GCMI modules are proposed to constrain the local and global biases of path context induction in the Transformer encoder and the path aggregator, respectively, by a straightforward geometric rule. Extensive experiments on 32 real-world relation reasoning tasks demonstrate that our method significantly outperforms 8 state-of-the-art baselines in terms of AP and AUC.

computer science, information systems, artificial intelligence,engineering, electrical & electronic