Learning Complete Topology-Aware Correlations Between Relations for Inductive Link Prediction

TL;DR

The paper tackles inductive link prediction by addressing edge-level semantic correlations between relations, a gap in prior graph-level centered methods. It introduces TACO, which combines a Relational Correlation Network (RCN) operating on a Relational Correlation Graph with a graph-structure module that reasons over Complete Common Neighbor (CCN) subgraphs. By classifying relation-pair topologies into seven patterns and preserving complete reasoning paths through CCN/CCN+ subgraphs, TACO achieves superior performance on multiple inductive benchmarks and offers interpretable relation correlations. The approach balances graph-level context with edge-level semantics, enabling robust generalization to unseen entities and scalable application to large knowledge graphs like YAGO3-10. Overall, TACO advances inductive reasoning in knowledge graphs by unifying relational correlations with complete subgraph structures, delivering practical gains in accuracy and interpretability.

Abstract

Inductive link prediction -- where entities during training and inference stages can be different -- has shown great potential for completing evolving knowledge graphs in an entity-independent manner. Many popular methods mainly focus on modeling graph-level features, while the edge-level interactions -- especially the semantic correlations between relations -- have been less explored. However, we notice a desirable property of semantic correlations between relations is that they are inherently edge-level and entity-independent. This implies the great potential of the semantic correlations for the entity-independent inductive link prediction task. Inspired by this observation, we propose a novel subgraph-based method, namely TACO, to model Topology-Aware COrrelations between relations that are highly correlated to their topological structures within subgraphs. Specifically, we prove that semantic correlations between any two relations can be categorized into seven topological patterns, and then proposes Relational Correlation Network (RCN) to learn the importance of each pattern. To further exploit the potential of RCN, we propose Complete Common Neighbor induced subgraph that can effectively preserve complete topological patterns within the subgraph. Extensive experiments demonstrate that TACO effectively unifies the graph-level information and edge-level interactions to jointly perform reasoning, leading to a superior performance over existing state-of-the-art methods for the inductive link prediction task.

Figures (6)

• Figure 1: An example of enclosing subgraph extraction process. Enclosing subgraph contains links between nodes in the $n$-hop neighborhood intersection of target nodes.
• Figure 2: An example in knowledge graphs.
• Figure 3: An overview of TACO. TACO consists of two modules: the relational correlation module and the graph structure module. We use a scoring network to score a triple based on the output of the two modules.
• Figure 4: An illustration of the topological patterns between any two relations and the corresponding RCGs. For the topological pattern where two relations are not connected, its corresponding RCG consists of two isolated nodes.
• Figure 5: Comparison between the $2$-hop enclosing subgraph method, $2$-hop CCN method, and $2$-hop CCN+ method in the same original knowledge graph.

Theorems & Definitions (6)

• Theorem 1
• proof
• Theorem 2
• proof
• Theorem 3
• proof