Approaches Used in Knowledge Representation

There is no single "correct" way to represent knowledge inside a machine — different approaches trade off expressiveness, ease of reasoning, and ease of construction differently.

1. Logical (Formal) Representation

Uses propositional or predicate logic to state facts as formulas that can be manipulated with formal inference rules (e.g., modus ponens). Precise and mathematically sound; covered in depth in the next lesson.

2. Semantic Networks

A graph of nodes (objects/concepts) connected by labelled edges (relationships) such as IS-A and HAS-A.

Inheritance lets "Canary" automatically acquire properties of "Bird" and "Animal" (e.g., can fly, needs food) without stating them explicitly.

3. Frames

A frame is a data structure bundling a concept together with its attributes ("slots") and their values, similar to an object in OOP.

Frame: Car
  slot: has-wheels     value: 4
  slot: powered-by      value: engine
  slot: is-a            value: vehicle
  slot: max-speed        default: 180 km/h

Frames support default values (assumed true unless overridden) and inheritance through IS-A slots, and can attach small procedures ("procedural attachments") that run when a slot is read or written.

4. Scripts

A script represents a stereotyped sequence of events for a common situation — useful for understanding narrative/text.

Example — "Restaurant script": enter, be seated, read menu, order, eat, pay bill, leave. Once a system recognises "restaurant," it can fill in unstated but expected steps (e.g., infer a menu was read even if the text never says so).

5. Production Rules

Knowledge stated as IF–THEN rules (already introduced as the engine of production systems in Unit 1) — e.g., "IF patient has fever AND rash THEN consider measles."

Comparing the Approaches