Towards Data-and Knowledge-Driven Artificial Intelligence: A Survey on Neuro-Symbolic Computing

TL;DR

Neural-symbolic computing (NeSy) aims to fuse symbolic reasoning with neural learning to address the interpretability and data-efficiency gaps of contemporary AI. The paper presents a four-dimensional taxonomy (neural-symbolic integration, knowledge representation, knowledge embedding, functionality) and outlines six integration types, detailing how symbolic knowledge can be embedded at data, sub-symbolic, architecture, or inference levels. It surveys a wide range of applications (scientific discovery, programming, QA, vision-language tasks, robotics, and math reasoning) and provides benchmarking across retrosynthesis, visual parsing, and math word problems, showing NeSy often outperforms purely neural approaches. The authors discuss open challenges, including scalability, compositional generalization, and knowledge acquisition, and argue for continued exploration of big-model integration to realize robust, explainable, and data-efficient AI.

Abstract

Neural-symbolic computing (NeSy), which pursues the integration of the symbolic and statistical paradigms of cognition, has been an active research area of Artificial Intelligence (AI) for many years. As NeSy shows promise of reconciling the advantages of reasoning and interpretability of symbolic representation and robust learning in neural networks, it may serve as a catalyst for the next generation of AI. In the present paper, we provide a systematic overview of the recent developments and important contributions of NeSy research. Firstly, we introduce study history of this area, covering early work and foundations. We further discuss background concepts and identify key driving factors behind the development of NeSy. Afterward, we categorize recent landmark approaches along several main characteristics that underline this research paradigm, including neural-symbolic integration, knowledge representation, knowledge embedding, and functionality. Next, we briefly discuss the successful application of modern NeSy approaches in several domains. Then, we benchmark several NeSy methods on three representative application tasks. Finally, we identify the open problems together with potential future research directions. This survey is expected to help new researchers enter this rapidly evolving field and accelerate the progress towards data-and knowledge-driven AI.

Figures (10)

• Figure 1: Structure of the overall review.
• Figure 2: Our four-dimensional taxonomy for NeSy. The dimensions are independent and not mutually exclusive, as explained in Sec. \ref{['subsec41']}. Note that the axes indicate the order in which each section is structured, not the sequence of their evolution. For example, things can be represented by First-Order Logic that Programming Language may not be able to express, and vice versa.
• Figure 3: Type 1: Symbolic Neuro Symbolic (also referred to as Neural Networks with Symbolic Input/Output).
• Figure 4: Type 2: Symbolic[Neuro] (also referred to as Neuro Subroutines).
• Figure 5: Type 3: Neuro$\vert$Symbolic (also referred to as Neural Learning + Symbolic Solver.)