Page 91 - DCAP506_ARTIFICIAL_INTELLIGENCE

Unit 6: Knowledge Representation using Rules

Matching among current state and the precondition of the rules is discussed as below. Notes
 Indexing
 A large number of rules  too sluggish to locate a rule
 Indexing: Use the existing state as an index into rules and choose the matching ones
instantly
 Only functions when preconditions of rules match accurately
 Only functions when preconditions of rules match exact board configuration

 It’s not forever apparent whether a rule’s preconditions are pleased by a specific
state
 There’s a swapping among the ease of writing rules (high-level descriptions) and
the ease of the matching process
 Matching with variables
 Generality in the statements of the rules: Require a search process to determine a match
among a particular state and the preconditions of a specified rule.
 Backward-chaining systems
 One-one matching algorithm

 Unification procedure + Depth-first backtracking to select individual rules
 Forward-chaining systems
 Many-many matching algorithm: RETE

 Approximate matching
 Rules should be applied if their preconditions roughly match the present situation

Example: A speech-understanding program
 Rules: An explanation of a physical waveform to phones (a, e, …)
 Physical signal: differences in the way individuals speak, result of background
noise, etc.

 Conflict resolution:
 Preferences dependent on rules:
 Specificity of rules

 Physical order of rules
 Preferences dependent on objects:
 Importance of objects
 Position of objects

 Preferences dependent on states:
 Assessment of states

LOVELY PROFESSIONAL UNIVERSITY 85