Knowledge Representation Overview

Principles

Definition

Knowledge representation is an interdisciplinary field in artificial intelligence concerned with designing formalisms and methodologies for modeling, organizing, and encoding information and knowledge computationally. It investigates how to structure knowledge in a way that enables automated reasoning, inference, and intelligent problem-solving by computer systems.

Ontological type

Core Formalisms

Reasoning Mechanisms

Applications in Cognition

Key developments

Key Figures

Symbolic Knowledge Codification era

Marvin Minsky helped popularize frame-based representations, organizing tacit knowledge into structured slots and fillers that supported explanation and goal-directed reasoning. Roger Schank advanced symbolic encoding of episodic knowledge through scripts and conceptual dependencies, enabling narrative understanding and analogical reasoning. Charles Forgy pioneered production-rule systems and the OPS family, demonstrating systematic, inspectable inference for knowledge engineering. Allen Newell articulated the knowledge level as a high-level abstraction for intelligent behavior, promoting principled representations that could be planned, explained, and updated.

Ontology and Semantic Web era

Susan Dumais[1] is associated with institutions such as the University of Colorado Boulder[3] and ETH Zurich[4] during this era. Her influential 1997 paper A solution to Plato's problem: The latent semantic analysis theory of acquisition, induction, and representation of knowledge[7] introduced LSA as a statistical, representation-based approach to knowledge acquisition and meaning extraction, providing a foundational method for bridging language and representations central to information retrieval and semantic interpretation in early ontology-relevant research. Moshe Y. Vardi[2] is affiliated with the Massachusetts Institute of Technology[5] and Stanford University[6] during this era. His 1994 paper Reasoning about knowledge[8] advanced formal methods for reasoning about knowledge, influencing the design of logical frameworks and automated reasoning essential for knowledge-based systems and the development of principled approaches in ontology engineering.

Neural-Language Graph Fusion era

Sebastian Riedel [1] is associated with the University of Washington [3] and the University College London [4] in this era. His key contribution is the Convolutional 2D Knowledge Graph Embeddings [6], a work that advanced neural-symbolic integration by applying convolutional architectures to graph embeddings to enable scalable, differentiable reasoning over knowledge graphs in this era. Pontus Stenetorp [2] is affiliated with the University College London [4] and Carnegie Mellon University [5] during this era. His contribution centers on the same framework, Convolutional 2D Knowledge Graph Embeddings [6], demonstrating how 2D convolutional reasoning can capture relational structure in graphs and catalyze neural-symbolic fusion for knowledge representation in this era [6].