Skip to main content

Knowledge Based Systems

  • Humans use not only methods, but also domain specific knowledge, in problem solving.
  • Top down approach

A system which is built around a knowledge base (not like a database) i.e. a collection of knowledge, taken from a human, and stored in such a way that the system can reason with it

  • Substitute, the certain extend, the experts.
  • Make good decisions, quickly, to come with problems
  • Domain specific: a narrow range of knowledge
  1. Acquire Knowledge - Transfer expertise from knowledge source to a program
  2. Represent knowledge - Symbolic encoding of propositions
  3. Apply the knowledge (reasoning) - Deduce logical consequences: to produce new propositions.

Knowledge acquisition/engineering​

  • Knowledge Engineering: Design, build, install a knowledge based system
  • Also called knowledge acquisition

Bottlenecks​

  • Human knowledge is complex and unstructured
  • Expert is often too close to the problem under consideration, difficult in seeing it objectively
  • Situation is worse when the knowledge source comprises of several experts

Stages​

  1. Define task
  2. Build domain Vocabulary (words, phrases, formulae)
  3. Develop a Model of the reasoning and how it is applied (paper exercises, flowcharts etc)

Knowledge Representation

Problems with natural language: Often ambiguous, syntax and semantics not fully understood.

Predicate Logic: Formal language to represent knowledge

  • Symbolic AI: Expressive, powerful to derive new knowledge from old, through mathematical deduction

Inference: infer conclusions from known statements.

Resolution Rule​

Resolution Rule: An inference method. For automated theorem proving, Proves new terms by contradictions

Input: a knowledge base and a statement

  1. Transfer knowledge base into Conjunctive Normal Form (CNF)
    • CNF: Propositions represented as a conjunction of clauses
  2. Negate the statement, add it to the knowledge base
  3. Unification: replace variables by a concrete instance (constant)
  4. Resolution: if a contradiction exists then the negated statement is false (applied repeatedly)
  5. Therefore, the original statement must be true