Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

A short Survey: Exploring knowledge graph-based neural-symbolic system from application perspective

Shenzhe Zhu, Shengxiang Sun

TL;DR

The paper addresses the interpretability gap in deep neural networks by surveying knowledge-graph-based neural-symbolic integration from application perspectives. It structures the landscape into three categories—Symbol for Neural, Neural for Symbol, and Hybrid Neural-Symbolic Integration—and reviews representative KG-driven methods for each, including KG-based recommender systems, Q&A, zero-shot/few-shot learning, and knowledge-enhanced language models. Notable hybrid approaches like CogQA, JointGT, HGNN-EA, and KIG are highlighted for their iterative, cyclic learning that fuses neural and symbolic components. The synthesis provides a roadmap for building more interpretable and efficient AI systems that leverage KG-structured knowledge to guide learning and reasoning, with future directions in multimodal learning, reasoning efficiency, and graph-integrated Transformer architectures.

Abstract

Advancements in Artificial Intelligence (AI) and deep neural networks have driven significant progress in vision and text processing. However, achieving human-like reasoning and interpretability in AI systems remains a substantial challenge. The Neural-Symbolic paradigm, which integrates neural networks with symbolic systems, presents a promising pathway toward more interpretable AI. Within this paradigm, Knowledge Graphs (KG) are crucial, offering a structured and dynamic method for representing knowledge through interconnected entities and relationships, typically as triples (subject, predicate, object). This paper explores recent advancements in neural-symbolic integration based on KG, examining how it supports integration in three categories: enhancing the reasoning and interpretability of neural networks with symbolic knowledge (Symbol for Neural), refining the completeness and accuracy of symbolic systems via neural network methodologies (Neural for Symbol), and facilitating their combined application in Hybrid Neural-Symbolic Integration. It highlights current trends and proposes future research directions in Neural-Symbolic AI.

A short Survey: Exploring knowledge graph-based neural-symbolic system from application perspective

TL;DR

The paper addresses the interpretability gap in deep neural networks by surveying knowledge-graph-based neural-symbolic integration from application perspectives. It structures the landscape into three categories—Symbol for Neural, Neural for Symbol, and Hybrid Neural-Symbolic Integration—and reviews representative KG-driven methods for each, including KG-based recommender systems, Q&A, zero-shot/few-shot learning, and knowledge-enhanced language models. Notable hybrid approaches like CogQA, JointGT, HGNN-EA, and KIG are highlighted for their iterative, cyclic learning that fuses neural and symbolic components. The synthesis provides a roadmap for building more interpretable and efficient AI systems that leverage KG-structured knowledge to guide learning and reasoning, with future directions in multimodal learning, reasoning efficiency, and graph-integrated Transformer architectures.

Abstract

Advancements in Artificial Intelligence (AI) and deep neural networks have driven significant progress in vision and text processing. However, achieving human-like reasoning and interpretability in AI systems remains a substantial challenge. The Neural-Symbolic paradigm, which integrates neural networks with symbolic systems, presents a promising pathway toward more interpretable AI. Within this paradigm, Knowledge Graphs (KG) are crucial, offering a structured and dynamic method for representing knowledge through interconnected entities and relationships, typically as triples (subject, predicate, object). This paper explores recent advancements in neural-symbolic integration based on KG, examining how it supports integration in three categories: enhancing the reasoning and interpretability of neural networks with symbolic knowledge (Symbol for Neural), refining the completeness and accuracy of symbolic systems via neural network methodologies (Neural for Symbol), and facilitating their combined application in Hybrid Neural-Symbolic Integration. It highlights current trends and proposes future research directions in Neural-Symbolic AI.
Paper Structure (23 sections, 8 equations, 11 figures, 2 tables)

This paper contains 23 sections, 8 equations, 11 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Categorization of neural-symbolic systems
  3. Neural for symbol
  4. Symbol for neural
  5. Hybrid neural-symbolic integration
  6. Methods based on knowledge graph
  7. Neural for symbol
  8. KG-based recommender systems
  9. KG-based Q&A systems
  10. Symbol for neural
  11. KG-driven zero-shot and few-shot learning
  12. Knowledge-enhanced pre-trained language models
  13. Hybrid neural-symbolic integration
  14. Future Trends & Direction
  15. Multimodal and multidomain learning
  16. ...and 8 more sections

Figures (11)

  • Figure 1: This diagram depicts three neural-symbolic system interactions: the orange curve for "Symbol for Neural", the pink for "Neural for Symbol", and the grey bidirectional line for "Hybrid neural-symbolic integration", highlighting their distinct collaborative dynamics.
  • Figure 2: Neural for symbol
  • Figure 3: Symbol for neural
  • Figure 4: Hybrid neural-symbolic integration
  • Figure 5: A two-layer receptive field (green entities) of the blue entity in a KG
  • ...and 6 more figures