Knowledge Representation Flashcards
(24 cards)
What is intelligence achieved by (in AI)?
Symbol patterns representing significant aspects of the problem domain, operations on these patterns, and search to select a solution.
What is the core description of AI?
AI = Search + Knowledge.
What does KR stand for in AI?
Knowledge Representation.
Why is KR important in complex problems?
Because representing the knowledge becomes a challenge beyond simple states and operators.
What is Knowledge Representation?
A set of syntactic and semantic assumptions to describe a world.
What are the two main aspects of KR?
Syntax and Semantics.
Why is natural language unsuitable for KR?
It is ambiguous.
What is Representational Adequacy?
Ability to represent all kinds of knowledge including uncertain knowledge.
What is Inferential Adequacy?
Allows the derivation of new knowledge from existing knowledge.
What is Inferential Efficiency?
Includes info to guide inference mechanisms to a solution more quickly.
What is Acquisitional Efficiency?
Allows incorporation of new knowledge efficiently.
What is Notational Convenience?
The resulting expression should be easy to write and read.
What needs to be described in KR?
- Objects,
- Quantity and quality,
- Events and time,
- Meta-knowledge.
What should KR be able to do?
- Handle qualitative knowledge,
- Allow inference,
- Represent general and specific situations,
- Capture semantics,
- Support meta-level reasoning.
What is an inference mechanism?
A procedure that operates on KR to produce new knowledge.
What are two search strategies for inference?
Data-driven (bottom-up) and Goal-driven (top-down).
What is Deductive Reasoning?
From known facts and rules, derive conclusions.
What is Inductive Reasoning?
From specific examples derive general conclusions.
What is Abductive Reasoning?
From rules and observations, infer a plausible explanation.
What are Logic-based KR techniques?
- Propositional logic,
- First-order logic,
- Fuzzy logic,
- Modal logic,
- CTL (Computational Tree Logic)
What are structured representations in KR?
- Semantic networks,
- Conceptual dependency,
- Scripts,
- Conceptual graphs,
- Frames.
What are rule-based KR techniques?
If-Then Rules.
What are the steps of mathematical induction?
- Prove for n=1,
- Assume true for n=k,
- Prove true for n=k+1.
Why does mathematical induction work?
If it holds for n=1 and for k implies k+1, it holds for all n (deduction).
