A Diffusion Model for Inductive Knowledge Graph Completion

Knowledge graph completion (KGC) is dedicated to deducing absent facts within incomplete knowledge graphs (KGs). The majority of antecedent studies exclusively address the transductive scenario, wherein all entities are observed throughout the training process, which is impractical for the everyday reality with constantly emerging entities. Recent works that concern this issue mainly focus on representing emerging entities with seen neighbors while ignoring the semantics in the queried relations. In this paper, we introduce a model named DKGC, which incorporates two purposefully designed entity encoding modules to capture the local and global entity representations in a relation-specific manner and a sampling module with a diffusion process to explore the potential semantics further. Extensive experiments on benchmark datasets indicate the superiority of our model for inductive KGC.

I. Introduction

Knowledge graphs (KGs) are referred to as a collection that accumulates real-world facts, such as NELL [6] and Freebase [4]. Each fact is denoted as (subject, relation, object), and we abbreviate it as (s, r, o) for simplicity, where s and o are two entities connected by the directed relation r. In order to quantify those discrete facts to continuous space, there are several knowledge graph embedding (KGE) models being proposed, such as TransE [5] and Distmult [33]. The projected embeddings inherently reveal the unseen links among the seen entities, which is also known as transductive knowledge graph completion. However, as the world evolves, there are always emerging facts with unseen entities. We consider such a scenario as inductive knowledge graph completion.