Nexus Learn

Lecture 5: Knowledge Representation

Lecture Slides CS-511 Artificial Intelligence 8 min read Updated 8 Nov 2026

↓ Download slides (PowerPoint)

Recap of Lecture 4 and Homework Check

LAST LECTURE · 5 MIN

Agents

PEAS + environment

Agent types

Sensors → agent program → actuators. Rational = best expected performance.

Performance, Environment, Actuators, Sensors. Seven properties.

Reflex → model-based → goal → utility → learning.

Link: a model-based agent needs a model of the world. Today: how to WRITE that knowledge down.

Today's Plan (90 minutes)

ROADMAP

0-5

Recap of Lecture 4 and homework check

S02

5-8

Plan and learning outcomes

S03-S04

8-28

Module 1: Knowledge, knowledge-based agents, types of knowledge

S05-S09

28-38

Module 2: What makes a good representation?

S10-S12

38-56

Module 3: Semantic networks and inheritance (+ Python demo)

S13-S17

56-74

Module 4: Frames and production rules (+ Python demo)

S18-S22

74-84

Class activity: represent UAF Burewala knowledge

S23

84-90

Mistakes, recap quiz, homework

S24-S26

Learning Outcomes (SLOs)

BY THE END OF THIS LECTURE YOU WILL BE ABLE TO

SLO 5.1

Differentiate data, information and knowledge, and describe a knowledge-based agent (TELL / ASK)

→ CLO2 · C2 Understand

SLO 5.2

Classify knowledge as declarative, procedural, meta, heuristic or structural

→ CLO2 · C2 Understand

SLO 5.3

Judge a representation using the four properties of a good KR system

→ CLO2 · C2 Understand

SLO 5.4

Represent knowledge as a semantic network and use inheritance with exceptions

→ CLO2 · C3 Apply

SLO 5.5

Represent knowledge as frames and production rules, and trace forward chaining

→ CLO2 · C3 Apply

  • Course Learning Outcome(s) served
  • CLO2: Demonstrate first-order predicate calculus, logical reasoning and problem-solving using Prolog language.

Slide 5

MODULE 1

What Is Knowledge?

Data, information, knowledge - and agents that store and use it

Data, Information and Knowledge

MODULE 1 · KNOWLEDGE

↑ more useful for decisions

Knowledge

Information

Data

Intelligence Needs Knowledge

MODULE 1 · KNOWLEDGE

Voluminous

Hard to describe exactly

Always changing

Very large. All the facts about crops, diseases and medicines.

What is 'a little fever' or 'good soil'?

New crop varieties, new virus strains, new electricity tariffs.

Knowledge-Based Agents: TELL and ASK

MODULE 1 · KNOWLEDGE

Percept

TELL KB: what I see

ASK KB: what should I do?

TELL KB: I did action A

Action

  • Knowledge base (KB): a set of sentences (facts and rules) written in a representation language.
  • TELL: add a new sentence to the KB.
  • ASK: query the KB; the answer is INFERRED (worked out) from what was told.
  • Declarative approach: tell the agent what is true; let inference work out what to do.

Example: crop advisor

  • TELL: leaves have yellow stripes; weather is cool.
  • ASK: which disease?
  • KB answers: yellow (stripe) rust of wheat.

Types of Knowledge

MODULE 1 · KNOWLEDGE

Quick rule: declarative = WHAT, procedural = HOW, meta = ABOUT knowledge, heuristic = RULE OF THUMB, structural = RELATIONS.

Slide 10

MODULE 2

What Makes a Good Representation?

Four properties to judge any knowledge representation

Four Properties of a Good KR System

MODULE 2 · GOOD REPRESENTATION

Representational adequacy

Inferential adequacy

  • Can it express ALL the knowledge needed?
  • Excel cannot easily say 'every crop needs water'.
  • Can it derive NEW knowledge from old?
  • Cotton is a crop + crops need water → cotton needs water.

Inferential efficiency

Acquisitional efficiency

  • Can it derive answers QUICKLY?
  • An IS-A hierarchy finds inherited facts fast.
  • Can new knowledge be ADDED easily?
  • Add a new crop by one IS-A link.

Truth: no single method is best at all four. That is why AI uses several representations.

Main Knowledge Representation Methods

MODULE 2 · GOOD REPRESENTATION

Logic

Semantic networks

Frames

Production rules

  • Exact sentences with true/false meaning.
  • 'All crops need water.' → L06-L09
  • Nodes (concepts) joined by labelled links (IS-A, HAS).
  • Google Knowledge Graph
  • A record with slots and fillers for a typical object.
  • A patient form
  • IF conditions THEN action.
  • Expert systems (L19)

Today: semantic networks, frames and production rules. Logic gets the next three lectures.

Slide 13

MODULE 3

Semantic Networks

Knowledge as a graph of concepts and relations

A Semantic Network

MODULE 3 · SEMANTIC NETWORKS

breathes

DOES

NEEDS

Animal

Crop

water

IS-A

feathers

IS-A

IS-A

HAS

Bird

Kharif crop

Rabi crop

can fly

CAN

IS-A

IS-A

IS-A

IS-A

Sparrow

Ostrich

Cotton

Wheat

Read it: nodes = concepts; labelled arrows = relations. IS-A always points from the specific to the general.

Inheritance and Exceptions

MODULE 3 · SEMANTIC NETWORKS

Inheritance: store a fact ONCE at the most general level; every specific node below gets it automatically.

Exception rule: the MOST SPECIFIC value wins (Ostrich's 'cannot fly' beats Bird's 'can fly').

Demo: Inheritance in Python

MODULE 3 · SEMANTIC NETWORKS

SEMANTIC NETWORK (FULL CODE: NOTES §3.3)

REAL OUTPUT

  • isa = {"Sparrow": "Bird", "Ostrich": "Bird", "Bird": "Animal", ...}
  • props = {"Bird": {"can_fly": True, ...},
  • "Ostrich": {"can_fly": False}, ...}
  • def lookup(node, prop):
  • path = [node]
  • while node is not None:
  • if prop in props.get(node, {}):
  • return props[node][prop], path
  • node = isa.get(node) # climb IS-A
  • if node: path.append(node)
  • return None, path
  • Sparrow can_fly = True [Sparrow>Bird]
  • Ostrich can_fly = False [Ostrich]
  • Ostrich breathes = True [Ostrich>Bird>Animal]
  • Cotton sown_in = summer (Apr-Jun) [Cotton>Kharif crop]
  • Cotton needs_water = True [Cotton>Kharif crop>Crop]
  • Wheat product = None [Wheat>Rabi crop>Crop]

Notice: 'Wheat product = None' - a KB knows only what it has been told.

Semantic Networks: Strengths and Weaknesses

MODULE 3 · SEMANTIC NETWORKS

Next: logic gives exact meaning for NOT, OR, ALL and SOME (Lectures 6-7).

Slide 18

MODULE 4

Frames and Production Rules

Records with slots, and IF-THEN knowledge

Frames: Slots and Fillers

MODULE 4 · FRAMES

Like a form

  • A hospital admission form: fixed slots, the patient's values. A default fills empty slots.
  • In OOP: class = generic frame, object = instance.

Frame: a record for a typical object: slots (attributes) + fillers (values, defaults, procedures).

Frame Hierarchy, Defaults and Instances

MODULE 4 · FRAMES

Student (generic)

BSCS-5th (generic)

Hira (instance)

Inherited: Hira's university (UAF) and campus come from the frames above. Her CGPA is her own.

OOP link: generic frame ≈ class; instance frame ≈ object; inheritance works the same way.

Production Rules and Forward Chaining

MODULE 4 · PRODUCTION RULES

RULES + WORKING MEMORY (FULL CODE: NOTES §4.3)

REAL OUTPUT

  • rules = [
  • ({"fever","body_ache"}, "viral_infection_possible"),
  • ({"viral_infection_possible","dengue_season"},
  • "test_for_dengue"),
  • ({"test_for_dengue","low_platelets"}, "dengue_suspected"),
  • ({"dengue_suspected"}, "refer_to_hospital"),
  • ]
  • facts = {"fever","body_ache","dengue_season","low_platelets"}
  • # repeat: fire every rule whose IF-part is in facts
  • Start facts: ['body_ache', 'dengue_season', 'fever', 'low_platelets']
  • Cycle 1: IF body_ache AND fever THEN viral_infection_possible
  • Cycle 2: IF dengue_season AND viral_infection_possible THEN test_for_dengue
  • Cycle 3: IF low_platelets AND test_for_dengue THEN dengue_suspected
  • Cycle 4: IF dengue_suspected THEN refer_to_hospital
  • Facts in memory now: 8
  • Final conclusion: refer_to_hospital

Forward chaining: start from facts, fire rules, add conclusions until nothing new.

Comparing the Representations

MODULE 4 · SUMMARY

Choose by task: relations → semantic net; records → frames; expert IF-THEN advice → rules; exact proof → logic.

Activity: Represent UAF Burewala Knowledge

CLASS ACTIVITY · GROUPS OF 3 · 10 MIN

Groups of 3. Use these facts: 'UAF Burewala is a sub campus of UAF. BSCS is a degree offered at UAF Burewala. Ali is a BSCS student. Every UAF student needs 75% attendance. BSCS students study CS-511.'

Draw a semantic network with at least 5 nodes and IS-A / other links

Write an instance frame for Ali

Using inheritance, answer: does Ali need 75% attendance? Show the path

Write 2 production rules about attendance and exam eligibility

Write a generic frame 'BSCS-Student' with 5 slots (one default)

Classify each fact as declarative, procedural, structural, meta or heuristic

Common Mistakes

WATCH OUT

WRONG IDEA

CORRECT IDEA

✗ Data and knowledge are the same.

✓ Knowledge = information + rules/experience that can be used to decide.

✗ IS-A arrows point from general to specific.

✓ IS-A points from the specific (Sparrow) to the general (Bird).

✗ Every fact must be stored at every node.

✓ Store it once at the most general level; inheritance passes it down.

✗ A default value can never change.

✓ A default is used only until a specific value overrides it (ostrich).

✗ Forward chaining starts from the goal.

✓ Forward chaining starts from FACTS; backward chaining starts from the goal.

Recap Quiz

CHECK YOUR UNDERSTANDING · 5 MIN

  • Q1. 'Patient's temperature is 38 °C' is an example of:
  • (a) Data (b) Information (c) Wisdom (d) Meta-knowledge
  • Q2. Knowing HOW to mix a spray is:
  • (a) Declarative (b) Procedural (c) Meta (d) Structural
  • Q3. 'Can the KR derive new facts from old ones?' is:
  • (a) Representational adequacy (b) Inferential adequacy (c) Acquisitional efficiency (d) Heuristic
  • Q4. Cotton IS-A Kharif crop IS-A Crop; Crop needs water. Cotton needs water by:
  • (a) Exception (b) Inheritance (c) Default (d) Forward chaining
  • Q5. In a frame, a value used when nothing else is known is a:
  • (a) Slot (b) Default (c) Instance (d) Rule
  • Q6. Firing rules from known facts until nothing new is added is:
  • (a) Backward chaining (b) Forward chaining (c) Inheritance (d) Search

Homework and Next Lecture

BEFORE THE NEXT CLASS

  • Task 1: Draw a semantic network for 'vehicles': Car, Bus, Rickshaw, Motorbike, Vehicle, Engine, Wheels - with IS-A and HAS links. Add one exception.
  • Task 2: Write a generic frame 'Mobile Phone' (6 slots, 2 defaults) and an instance frame for your own phone.
  • Task 3: Write 4 production rules for a simple tubewell controller (soil moisture, rain forecast, load-shedding) and trace forward chaining for one case.
  • Read: Kaplan Ch 2; Luger Ch 7.

NEXT LECTURE

  • Lecture 6:
  • Propositional Logic

← All CS-511 material