Knowledge representation

• MIT’s ConceptNet : open, common-sense knowledge
• Wikimedia’s Wikidata
• Google:
  • Knowledge graph : traditional approach, in production (API )
  • Knowledge vault : ML approach, experimental (paper)
    • Automated, with posterior probabilities of fact correctness
    • Prior knowledge mostly from existing KBs
• Metacademy : knowledge graphs for human learning (e.g., deep belief nets )

The Semantic Web

• RDF: Resource Description Framework
  • Primer and concepts
  • RDF Schema : extends RDF with classes and properties
• OWL: Web Ontology Language
  • A description logic
  • Primer and quick reference
• Examples: schema.org

Books

• Russell & Norvig, 2021, AI: A Modern Approach, 4th ed. (website ), Ch. 10: Knowledge representation
• Brachman & Levesque, 2004: Knowledge Representation and Reasoning
• Sowa, 2000: Knowledge Representation (website )
  • Greater than usual historical and philosophical emphasis
• Fagin, Halpern, Moses, Vardi, 1995: Reasoning about Knowledge
  • Halpern, 1986: Reasoning about knowledge: An overview (doi)

Frames

Frames , an early approach to KR, inspired object-oriented programming. Discussants at LtU have wondered whether there is even a meaningful difference between the two paradigms.

• Minsky, 1974: A framework for representing knowledge (pdf, online )
  • The original paper on frames
• Hayes, 1980: The logic of frames (pdf)
  • Influential criticism of frames in AI
  • “Most of ‘frames’ is just a new syntax for parts of first-order logic”

Knowledge graphs

“Knowledge graphs” is a catch-all term for graph-based KR systems. They can be implemented using graph databases and often lack a formal mathematical foundation.

• Hogan et al, 2020: Knowledge graphs (arxiv, HN )
  • Long survey paper on knowledge graphs
  • Mostly informal and example-driven, with a few formal definitions in Appendix B

Description logic

See page on description logic in the logic section.