Definition
Knowledge representation (KR) is the AI subfield concerned with how a system stores and reasons about information about the world. Major frameworks: logical formalism (predicate calculus, McCarthy), frames (structured templates with slots and defaults, Minsky), semantic networks (graphs of concepts and relations, Quillian), production rules (if-then, expert systems), and modern vector embeddings (continuous representations learned from data).
Why it matters
How it works
Choose a representation by analyzing the queries to be answered, the inferences to be drawn, and the data sources to be integrated. Logical formalisms excel at precise inference but explode combinatorially. Frames work for stereotyped scenarios but are hard to combine. Vector embeddings handle similarity and analogy but are opaque about logical structure. Hybrid approaches — neuro-symbolic systems — try to get the best of both.