Ch4: Knowledge Representation Issues

Q What is KR?
Q Representation and
mapping
Q Approaches to KR
Q Issues in KR
Q The Frame problem

Slide 1

What is KR?
Q R. Davis, H. Schrobe, P. Szolovits (1993):
1. A surrogate
2. A set of ontological commitments
3. A fragmentary theory of intelligent reasoning
4. A medium for efficient computation
5. A medium of human expressions

Slide 2

Representation and Mapping
In order to solve the complex problems encountered in AI,
one needs
– Large amount of Knowledge
– Mechanisms for manipulating that knowledge to create
solutions to new problems.
Kinds of entities
– Facts: things we want to represent (knowledge level)
Representation mapping
– Representations of facts in some chosen formalism:
things we can manipulate (symbol level)
Slide 3

Representation and Mapping
Structuring these entities is as two levels:
1.The Knowledge level – at which facts including the agent’s
behaviors and current goals are described.
2.The Symbol Level – at which representations of objects at
the knowledge level are defined in terms of symbols that
can be manipulated by programs.
The Links called representation mappings focus on
facts, on representations and on the two way mappings
that must exist between them.
The forward representation mapping maps from facts
to representations.
The backward representation mapping goes the
other way from representations to facts.
Slide 4

Representation and Mapping
Facts

Internal
Representations
English
understanding

Reasoning
programs

English
generation

English
Representations

Mappings between Facts
and Representations
Slide 5

Representation and Mapping
An example using mathematical logic as the
representational formalism:
– Spot is a dog:
the fact represented by that English sentence can be
represented in logic as
dog(Spot)
– Every dog has a tail:
the logical representation of the fact is given by
x: dog(x)  hastail(x)
– Using the deductive mechanisms of logic:
hastail(Spot)
Using an appropriate backward mapping function,the English
sentence is given as
Spot has a tail

Slide 6

Representation and Mapping
Q Fact-representation mapping is not one-to-one
– Every dog has a tail

vs. All dogs have tails

Q Good representation can make a reasoning
program trivial
– The Mutilated Checkerboard Problem: “Consider a
normal checker board from which two squares, in
opposite corners, have been removed. The task is to
cover all the remaining squares exactly with
dominoes, each of which covers two squares. No
overlapping, either of dominoes on top of each other
or of dominoes over the boundary of the mutilated
board are allowed. Can this task be done?”
Slide 7

Representation and Mapping

No. black squares = 30
No. white squares = 32

(b)

(c)

(a)

Slide 8

Representation and Mapping
Initial

desired real reasoning

facts
forward
representation
mapping
Internal
representations
of initial facts

Final
facts

backward
representation
mapping

operation
of program

Internal
representations
of final facts

Slide 9

Representation and Mapping
– The dotted line represents the abstract reasoning process that a
program is intended to model.
– The solid line across the bottom represents the concrete
reasoning process that a particular program performs.
– This program successfully models the abstract process to the
extent that , when the backward representation mapping is
applied to the program’s output, the appropriate final facts are
actually generated.
– If either the program’s operations or one of the representation
mappings is not faithful to the problem being modeled, then the
final facts will probably not be the desired ones.
– If no good mapping can be defined for a problem,it will not be
able to produce answers that correspond to real answers to the
Slide
problem.
10

Approaches to KR
Q Four properties of a good presentation of knowledge in
a particular domain:
1. Representational adequacy the ability to
represent all of the kinds of knowledge that are
needed in the domain.
2. Inferential adequacy the ability to manipulate the
representational structures in such a way as to
derive new structures corresponding to new
knowledge inferred from old.
3. Inferential efficiency the ability to incorporate into
the knowledge structure additional information that
can be used to focus the attention of the inference
mechanisms in the most promising directions. Slide 11

Approaches to KR
4 . Acquisitional efficiency the ability to acquire new
information easily.the program itself would be able to
control knowledge acquisition.
Q No single system that optimizes all of the capabilities
for all kinds of knowledge.
=> Multiple techniques for KR exist.

Slide 12

Approaches to KR
Q Simple relational knowledge
– Represent declarative facts as a set of relations used
in database systems
– Provides very weak inferential capabilities
– May serve as the input to powerful inference engines

Slide 13

Approaches to KR
Q Inheritable knowledge
– Objects are organized into classes and classes
are organized in a generalization hierarchy
– Inheritance is a powerful form of inference, but not
adequate
Mammal
isa

Person
Red

uniform
color

has-part

Nose

instance

Owen

team

Liverpool
Slide 14

Approaches to KR
Q Inferential knowledge
– Facts represented in a logical form (e.g. First-Order
Logic: FOL), which facilitates reasoning.
– An inference engine is required.

Slide 15

Approaches to KR
Q Procedural knowledge
– Representation of “how to make it” rather than “what it
is”
– Procedural knowledge can be represented in
programs in many ways:
• Code in some programming language,
such as Lisp
→May have inferential efficiency, no inferential
adequacy (difficult to write a program that can
reason about another program’s behaviour),
acquisitional efficiency (b/c of the process of
updating and debugging large pieces of code)
Slide 16

Approaches to KR
• Procedural knowledge as production rules
→Distinction between declarative and procedural
knowledge is difficult
If:
white pawn at square(file e, rank 2) AND
square(file e, rank 3) is empty AND
square(file e, rank 4) is empty
Then:
move pawn from square(file e, rank 2) to
square(file e, rank 4).

Slide 17

Issues in KR
Q Choosing the Granularity
– High-level facts may not be adequate for inference.
Substantial work must be done to reduce the
knowledge into primitive form.
– Low-level primitives may require a lot of
storage.

Slide 18

Homework
Q Reading
R. Davis, H. Schrobe, P. Szolovits (1993): “What is a
knowledge representation?”

Slide 19