HoGRN: Explainable Sparse Knowledge Graph Completion via High-Order Graph Reasoning Network | IEEE Journals & Magazine | IEEE Xplore

HoGRN: Explainable Sparse Knowledge Graph Completion via High-Order Graph Reasoning Network


Abstract:

Knowledge Graphs (KGs) are becoming increasingly essential infrastructures in many applications while suffering from incompleteness issues. The KG Completion (KGC) task a...Show More

Abstract:

Knowledge Graphs (KGs) are becoming increasingly essential infrastructures in many applications while suffering from incompleteness issues. The KG Completion (KGC) task automatically predicts missing facts based on an incomplete KG. However, existing methods perform unsatisfactorily in real-world scenarios. On the one hand, their performance will dramatically degrade along with the increasing sparsity of KGs. On the other hand, the inference procedure for prediction is an untrustworthy black box. This paper proposes a novel explainable model for sparse KGC, compositing high-order reasoning into a Graph Convolutional Network (GCN), namely HoGRN. It can not only improve the generalization ability to mitigate the information insufficiency issue but also provide interpretability while maintaining the model's effectiveness and efficiency. Two main components are seamlessly integrated for joint optimization. First, the high-order reasoning component learns high-quality relation representations by capturing endogenous correlation among relations. This can reflect logical rules to justify a broader range of missing facts. Second, the entity updating component leverages a weight-free GCN to efficiently model KG structures with interpretability. For evaluation, we conduct extensive experiments–the results of HoGRN on several sparse KGs present considerable improvements. Further ablation and case studies demonstrate the effectiveness of the main components.
Published in: IEEE Transactions on Knowledge and Data Engineering ( Volume: 36, Issue: 12, December 2024)
Page(s): 8462 - 8475
Date of Publication: 16 July 2024

ISSN Information:

Funding Agency:


I. Introduction

Knowledge Graphs (KGs) are one of the most effective ways to organize world facts in the form of a directed graph, where nodes denote entities and edges denote their relations. Recently, a series of KGs have been curated in various domains, including medicine [1], health care [2], and finance [3]. They are playing an increasingly important role in a variety of applications, such as drug discovery [4], user modeling [5], [6], dialog system [7], and question answering [8], [9]. However, existing KGs suffer from serious incompleteness issues. Due to the high cost of manual labeling, KG Completion (KGC) becomes an essential task for predicting missing facts based on an incomplete KG.

References

References is not available for this document.