Knowledge Representation and Expert Systems

Question 1

STRUCTURE OF EXPERT SYSTEMS

Answer 1

An expert system behaves like an expert in some narrow area of expertise

An ES generally has capabilities to:

• solve problems using domain specific information (and perhaps with uncertainty).
• interact with the user to take input to answer questions and deliver explanations

Three main modules; knowledge base, inference engine and user interfase

Question 2

FEATURES OF EXPERT SYSTEMS

Answer 2

• Problem solving in the area of expertise
• Relying heavily on domain knowledge
• Interaction with user during and after problem solving
• Explanation: ability of explain results to user

Question 3

DESIRABLE FEATURES OF RULES

Answer 3

• Modularity: each rule defines a relatively independent piece of knowledge
• Incrementality: new rules added (relatively independently) of other rules
• Modifiable and transparent: can see what goes wrong or needs addition, has
  explicit rules.
• Can represent uncertainty
• Chaining of rules

Question 4

KNOWLEDGE BASE DEVELOPMENT IN A DOMAIN

Answer 4

The facts and rules of a KB in a domain are derived from experts in a process of knowledge elicitation, which is basically a careful and systematic analysis of what domain experts know and do

Question 5

EXPERT SYSTEMS V. MACHINE LEARNING

Answer 5

Expert systems (as outlined previously) are markedly different from
statistical/machine learning approaches.
* Scale of data
* Questions asked and answered
* Inference v. classification (though classification can be handled via rules and
inference)
* Representation of human knowledge v representation of data
* Development process
* Different tools for different data and purposes

Question 6

FORWARD VS. BACKWARD CHAINING

Answer 6

data : goals
evidence : hypotheses
findings, observations : explanations, diagnoses
manifestations : diagnoses, cause

• Backward chaining: “goal driven”, e.g. from diagnosis to findings
• Forward chaining: “data driven”, e.g. from findings to diagnosis