Abstract:Knowledge graph reasoning (KGR) aims to infer new knowledge or detect noises, which is essential for improving the quality of knowledge graphs. Recently, various KGR techniques, such as symbolic- and embedding-based methods, have been proposed and shown strong reasoning ability. Symbolic-based reasoning methods infer missing triples according to predefined rules or ontologies. Although rules and axioms have proven effective, it is difficult to obtain them. Embedding-based reasoning methods represent entities and relations as vectors, and complete KGs via vector computation. However, they mainly rely on structural information and ignore implicit axiom information not predefined in KGs but can be reflected in data. That is, each correct triple is also a logically consistent triple and satisfies all axioms. In this paper, we propose a novel NeuRal Axiom Network (NeuRAN) framework that combines explicit structural and implicit axiom information without introducing additional ontologies. Specifically, the framework consists of a KG embedding module that preserves the semantics of triples and five axiom modules that encode five kinds of implicit axioms. These axioms correspond to five typical object property expression axioms defined in OWL2, including ObjectPropertyDomain, ObjectPropertyRange, DisjointObjectProperties, IrreflexiveObjectProperty and AsymmetricObjectProperty. The KG embedding module and axiom modules compute the scores that the triple conforms to the semantics and the corresponding axioms, respectively. Compared with KG embedding models and CKRL, our method achieves comparable performance on noise detection and triple classification and achieves significant performance on link prediction. Compared with TransE and TransH, our method improves the link prediction performance on the Hits@1 metric by 22.0% and 20.8% on WN18RR-10% dataset, respectively.

Difference-Aware Iterative Reasoning Network for Key Relation Detection.

Visual relationship detection with a deep convolutional relationship network

Iterative Visual Relationship Detection via Commonsense Knowledge Graph

Knowledge-Based Reasoning Network for Relation Detection

Neural Axiom Network for Knowledge Graph Reasoning

Graph Reasoning Networks for Visual Question Answering

Relation-Aware Multi-hop Reasoning forVisual Dialog

Hierarchical Semantic Enhanced Directional Graph Network for Visual Commonsense Reasoning

A Probabilistic Graphical Model Based on Neural-symbolic Reasoning for Visual Relationship Detection

RR-Net: Relation Reasoning for End-to-End Human-Object Interaction Detection

Attentive Relational Networks for Mapping Images to Scene Graphs

Detecting Visual Relationships with Deep Relational Networks

Relation-Aware Reasoning with Graph Convolutional Network.

ViRel: Unsupervised Visual Relations Discovery with Graph-level Analogy

RR-Net: Injecting Interactive Semantics in Human-Object Interaction Detection

Improving Rare Relation Inferring for Scene Graph Generation Using Bipartite Graph Network

Hierarchical Reasoning Network for Human-Object Interaction Detection

Illation of Video Visual Relation Detection Based on Graph Neural Network.

Joint Modeling of Visual Objects and Relations for Scene Graph Generation.

Hierarchical Graph Attention Network for Visual Relationship Detection

Relational graph neural network for situation recognition