Distributed Representations of Entities in Open-World Knowledge Graphs

TL;DR

This paper introduces Decentralized Attention Network (DAN), a novel GNN-based approach for knowledge graphs that effectively handles new entities in open-world scenarios through neighbor-based semantics and self-distillation.

Contribution

The paper proposes DAN, a new GNN model that distributes entity semantics among neighbors and employs self-distillation for training, addressing the challenge of unseen entities in open-world knowledge graphs.

Findings

DAN achieves competitive results on entity alignment and prediction tasks.

DAN significantly outperforms existing methods in open-world settings.

Theoretical analysis confirms the effectiveness of the approach.

Abstract

Graph neural network (GNN)-based methods have demonstrated remarkable performance in various knowledge graph (KG) tasks. However, most existing approaches rely on observing all entities during training, posing a challenge in real-world knowledge graphs where new entities emerge frequently. To address this limitation, we introduce Decentralized Attention Network (DAN). DAN leverages neighbor context as the query vector to score the neighbors of an entity, thereby distributing the entity semantics only among its neighbor embeddings. To effectively train a DAN, we introduce self-distillation, a technique that guides the network in generating desired representations. Theoretical analysis validates the effectiveness of our approach. We implement an end-to-end framework and conduct extensive experiments to evaluate our method, showcasing competitive performance on conventional entity alignment and entity prediction tasks. Furthermore, our method significantly outperforms existing methods in open-world settings.