- LPU MCA Semester 1
- CAB107
- Unit 5
CAB107 · Unit 5: Artificial intelligence techniques
Rule-based expert systems
10 practice questions · in the ETE syllabus
Key points
- Knowledge base plus inference engine are the core
- Forward chaining is data-driven, backward is goal-driven
- Explanation facility traces the rules that fired
- Rules are costly to build and do not learn
A rule-based expert system captures the knowledge of a human specialist as if-then rules to solve problems in a narrow domain, such as diagnosing faults.
Components:
| Component | Role |
|---|---|
| Knowledge base | Rules and facts supplied by experts |
| Working memory | Facts about the current case |
| Inference engine | Applies rules to derive conclusions |
| Explanation facility | Shows which rules fired, justifying a result |
| User interface | Dialogue with the user |
A knowledge engineer elicits knowledge from domain experts and encodes it as rules in the knowledge base.
Forward chaining starts from known facts and derives conclusions; backward chaining starts from a goal and works back to the facts that would support it.
Worked example: rules A and B give C, C gives D, D gives F. With only A and E known and B absent, F cannot be proven: C needs B, so D and F are never reached. Backward chaining from F fails at the same missing fact.
Strengths: transparent reasoning and good for clear if-then domains. Limitations: they do not learn automatically, and obtaining and maintaining rules from experts is slow and costly; they cope poorly with noisy or perceptual problems such as face recognition.
Exam trap: expert systems do not learn from data; the explanation facility is what lets doctors see why a diagnosis was made.
Try two questions
How does forward chaining differ from backward chaining?
- Forward chaining starts from known facts and derives conclusions, while backward chaining starts from a goal and works back to the facts
- Forward chaining starts from a goal and works back to the facts, while backward chaining starts from known facts and derives conclusions
- Forward chaining works only on numeric data, while backward chaining works only on text data and symbols
- Forward chaining needs no rules at all, while backward chaining needs a database of rules and facts
Show the answer
A. Forward chaining is data-driven; backward chaining is goal-driven.
Why not B: This swaps the two directions.
Why not C: Both operate over rules and facts of any kind.
Why not D: Both need rules.
Two rules can fire at the same time, and the system picks the one with more conditions in its premise. Which conflict resolution strategy is this?
- Specificity
- Recency
- Random selection
- Refraction
Show the answer
A. Specificity prefers the more specific rule over a general one.
Why not B: Recency prefers rules matching the newest facts.
Why not C: Random selection ignores the number of conditions.
Why not D: Refraction prevents the same rule firing twice on the same facts.