Knowledge Representation

Introduction

• Domain specific knowledge is needed to solve some problems.

• Knowledge base – representation.• Inference techniques• Use to prove facts.• Use to answer queries

Knowledge Representation Schemes

• Logical representation schemes – logical proofs, e.g. predicate logic.

• Procedural representation schemes – procedural list of instructions, e.g. production rule systems.

• Network representation schemes – graphical representation, e.g. semantic networks and conceptual graphs.

• Structured representation schemes –extension of network schemes, e.g. scripts and frames.

• Structures for incomplete or inconsistent knowledge: Bayesian reasoning, Dempster-Shafer, fuzzy logic.

.

Semantic Networks

• Consists of nodes and arcs.• Nodes represents objects.• Arcs represents relationships between the

objects.• Relationships:

– is-a – Instance– has/has-a– Others

.

Example 1Clyde is an elephant.

.

Mammal

Elephant

Clyde

is-a

instance

Example 2

Tail Dog 4 Legs

GreatDane Barks

Tall

Pluto Spots

has has

is-a communicates

heightinstance

has

Example 3

Mammal

Person Nose

Pee-Wee BrooklynBlue Reese Dodgers

is-a

has-a

instance

teamuniformcolour

Example 4

John’s height is 72.

John is taller than Bill.

John 72height

John

H1

Bill

H2

height height

greaterthan

Example 5John gives Mary the book.

Give

Book32

recipient

object

giverMary John

Using Case Frames

• Problems with semantic networks• Attempts to standardize semantic networks

led to the development of case frames.• Each sentence is represented by a case

frame.• Each case frame represents an actions.• Case frames define case relationships: agent,

object, instrument, location and time..

Example: Case Frame

glue

Fix

instrument

time

objectagentSarah

past

chair

ExercisesConstruct semantic networks for the following statements:• Pompeian(Marcus), Blacksmith(Marcus). • Mary gave the green coloured vase to

her favorite cousin.• John went downtown to deposit his

money in a bank.

Conceptual Graphs

• Is a connected bipartite graph.• Conceptual relation nodes represent

relations between concepts. • The arcs connecting nodes are not

labeled.• Concepts are represented by boxes.• Relations are represented by arcs.

Conceptual Graph Relations

bird flies

dog colour brown

child parents

mother

father

1-ary relation

2-ary relation

3-ary relation

ExampleMary gave John the book.

Maryagent

agent

give object

book

person:

person:John

Unique Markers

• If an individual object is unknown a unique marker can be used in place of the name of the object.

• A unique marker is comprised of a hash symbol (#) followed by a number.

• Each object has its own unique marker.

Example 1: Unique Markers

Emmacolour brown dog:

#1352colour brown dog:

Example 2: Unique Marker

name ”emma”

#1352colour brown dog:

Example Her name was McGill and she called herself Lil,

but everyone knew her as Nancy.

name ”Nancy”

#941name

“McGill”

person: name

“Lil”

Generic Markers

• A generic marker is used to represent an unspecified individual of a type.

• A generic marker is represented by an asterisk *.

• Name variables can also be used, e.g., *X to indicate an unspecified individual.

Example: Generic MarkerThe dog is scratching its ear with its paw.

objectagent scratch

instrument

dog: *X

paw

ear

dog: *X

part

part

Display FormJohn is going to Boston by bus.

agentperson:John go destination

city:Boston

instance

bus

Linear Form

[Go] –(Agnt) → [Person: John](Dest) → [City: Boston](Inst) → [Bus]

John is going to Boston by bus.

ExercisesConstruct conceptual graphs for the following statements:• Jane gave Tom an ice cream.• Basketball players are tall.• Paul cut down the tree with an axe.• Place all the ingredients in a bowl and

mix thoroughly.