Sie sind auf Seite 1von 17

Ch7: Structured Knowledge Representation

 Semantic nets
 Frames
 Conceptual graphs
 Conceptual dependency

Faculty of Computer Science and Engineering HCMUT

Slide 2

Semantic Nets
 The main ideas
For representing knowledge in structures
The meaning of a concept comes from the ways it is
connected to other concepts
Labelled nodes representing concepts are connected
by labelled arcs representing relations

Faculty of Computer Science and Engineering HCMUT

Slide 3

Semantic Nets
Mammal
isa

Person
Red

uniform
color

has-part

Nose

instance

Owen

team

Liverpool

Use inheritance: has-part(Owen, Nose)

Faculty of Computer Science and Engineering HCMUT

Slide 4

Semantic Nets
 Intersection Search (Quillian, 1968):
To find relationships among objects
Spreading activation out from each of two nodes and
seeing where the activation met
Mammal
isa

Person
Red

uniform
color

has-part

Nose

instance

Owen

team

Liverpool

What is the connection between Liverpool and Red?


Faculty of Computer Science and Engineering HCMUT

Slide 5

Semantic Nets
 Presenting the predicates:
Unary predicate
Binary predicate

Mammal
isa

Person
Red

uniform
color

has-part

Nose

instance

Owen

team

Liverpool

person(Owen) instance(Owen, Person)

Faculty of Computer Science and Engineering HCMUT

Slide 6

Semantic Nets
Three or more place predicate
John gives Mary a book
Give
John

agent

give(John, Mary, book)

instance
object

Book
instance

beneficiary

Mary

Faculty of Computer Science and Engineering HCMUT

Slide 7

Semantic Nets
 Making Some Important Distinctions
John

height

1.8

Bill

John

height

height
greater-than

H1
John
height
greater-than

H1

H2

Bill
height

H2

value

1.8
Faculty of Computer Science and Engineering HCMUT

Slide 8

Semantic Nets
 Partitioned Semantic Nets
The dog bit the mail-carrier
Dogs

Bite

instance

assailant

Mail-Carrier

instance

victim

Faculty of Computer Science and Engineering HCMUT

instance

Slide 9

Semantic Nets
Every dog has bitten a mail-carrier
GS

Dogs

instance

instance

form

Bite

Mail-Carrier

instance

assailant

instance
S1

victim

Every dog has bitten the mail-carrier in town


GS
instance

Dog
form

Bite

instance

Mail-Carrier

instance

b
assailant

victim

instance

S1

Faculty of Computer Science and Engineering HCMUT

Slide 10

Semantic Nets
Every dog has bitten every mail-carrier
Dogs

Bite

instance

Mail-Carriers

instance

instance

S1
assailant

GS

victim

instance

form

As we expand the range of problem-solving tasks


=> the representation becomes more complex
Faculty of Computer Science and Engineering HCMUT

Slide 11

Frame
 As we expand the range of problem-solving tasks, the
semantic-net representation becomes more complex
 Minsky, 1975 A Framework for Representing
Knowledge
 Knowledge is encoded in packets, called frames
 A Frame = A collection of attributes and associated
values (and possibly constraints on values)
 Frame system = collection of frames that are connected
to each other
 Attribute of one frame may be another frame
Faculty of Computer Science and Engineering HCMUT

Slide 12

Frame
 Example
(defineType Animals
(alive T)
(flies F)
)

(defineInstance
(type
(name
)

(defineType Mammals
(super-type Animals)
(legs
4)
)

(defineType Birds
(super-type Animals
(legs
2)
(flies
T)
)

Faculty of Computer Science and Engineering HCMUT

Tweety
Animals)
"Tweety")

Slide 13

Frame
(defineType Penguins
(super-type Birds
(flies
F)
)
(defineType
Cats
(super-type Mammals)
)
(defineType Bats
(super-type Mammals)
(legs
2)
(flies
T)
)

(defineInstance
(type
(name
(friend
)
(defineInstance
(type
(name
(friend
)
(defineInstance
(type
(name
)

Faculty of Computer Science and Engineering HCMUT

Opus
Penguins
"Opus")
Bill)
Bill
Cats)
"Bill")
Opus)
Pat
Bats
"Pat")
Slide 14

Frame
Animals
Alive
Flies

T
F

isa
Birds
Legs
Flies

Mammals

2
T

Legs

isa
Penguins
Flies

Cats

Bats

Legs
Flies

2
T

instance
Opus
Name
Friend

Opus

Bill
Name
Friend

Pat
Bill

Faculty of Computer Science and Engineering HCMUT

Name

Pat

Slide 15

Frame
(defineType Animals
(alive T)
(flies F)
)
Rel(Animals, alive, T)
Rel(Animals, flies, F)

(defineInstance
(type
(name
(friend
)

Opus
Penguins
"Opus")
Bill)

type(Opus, Penguins)
name(Opus, Opus)
friend(Opus, Bill)

Faculty of Computer Science and Engineering HCMUT

Slide 16

Frame
Animals
Alive
Flies

T
F

isa
Birds

Multiple
inheritance

Legs
Flies

Mammals

2
T

Legs

isa
Penguins
Flies

Cats

Bats

Legs
Flies

2
T

instance
Opus
Name
Friend

Opus

Bill
Name
Friend

Pat
Bill

Faculty of Computer Science and Engineering HCMUT

Name

Pat

Slide 17

Frame
Need an algorithm for traversing the isa hierarchy that
guarantees that specific knowledge will always
dominate more general facts.
Need an inheritance algorithm that reports the
ambiguity.

Faculty of Computer Science and Engineering HCMUT

Slide 18

Das könnte Ihnen auch gefallen