Chapter 6 Knowledge Representation
 

6.1 Introduction

                                                procedural knowledge
declerative knowledge  -----------------------> new declerative  knowledge

(x,y) --> declerative knowledge

IF ( X ? 4) THEN FILL X TO X = 4 ---> procedural knowledge

An object which does not contain given details is a simple object.
An object which can be classified according to given details is a structured object.

KR -----> knowledge formalized in a symbolic form
 

 

6.2 Knowledge Types

1. Procedural knowledge

               IF situation THEN action

               IF days sales outstanding > 30 THEN check involving procedure.
 

2. Declarative knowledge

               IF antecedent THEN consequent

               IF X is accounts payable THEN X belongs to curent liabilities.
                      conjunction (and) disjunction (or) of antecedents.

               IF (company ALPHA, loan????? application, loan size $4,500)
                    THEN (company ALPHA, reject Loan)          -too small explanation
 

6.3 Knowledge Representation Methods

1. Predicate Calculus
2. Frames
3. Semantic Networks
4. O-A-V Triplets
5. Production Rules
6. Neural Networks

 

6.3.1 Predicate Calculus

Elemantary unit in predicate calculus is an object. Statements about objects are called predicates. For example, is-blue(car) is an assertion that says that the cars is blue.

Predicate calculus describes a world which has the following form.

• There is a set of objects.
• There are a set of properties that each object may posses.
• There are functions. Given some objects each function will compute another object.
• There are relations which can hold between objects.

First-order Predicate Calculus       ===> Permits variables represent objects only.
Second-order Predicate Calculus  ===> Permits variables represent  predicates (functions)

6.3.2 Frames

• Provide a method for representing facts and relationships.

 
• Describes an object which contains slots for all the info associated with that object.

E.g.
 

6.3.3 Semantic Networks

A semantic net is a collection of objects called nodes. The nodes are connected together by arcs or links. Both the links and nodes are labeled.

Nodes : used to represent objects and descriptions. (Descriptors provide additional info about objects.)

Links :  relate objects and descriptors. Also represent relationships.

.

6.3.4 Object-Attribute-Value(O-A-V) Triplets

• A common way to represent factual information
• Used in MYCIN.

Attributes are general characteristics or properties associated with objects.For example, interest rate is an attribute for a bank loan.

O-A-V and Semantic Networks

O-A-V is a specialized case of the semantic-net approach.

        object      ==>  attribute         "has a"
        attribute  ?==  value              "is a"

The object ==> attribute link is a "has-a" link.
    A bank has a rate of interest.

The attribute ==>value links is a "is-a" link.
    12% is a rate of interest.
 

 

6.3.5 Production Rules

Rules are used to represent relationships. Rule-based knowledge representation employs
IF condition THEN action statements.
     (premise              (goal
     antecedent)          consequent)

For example,
 
              IF the heating element glows AND the bread is always dark
              THEN the toaster thermostat is broken

When the problem situation matches th IF part of a rule, the action specified by the THEN part of the rule is performed.
 

 

6.3.6 Neural Networks
 
Knowledge representation using Neural Networks  :

• forms a different approach to computation
• can be incorprated into silicon chips
• trained and not programmed
• specified by the topology, cell characteristics, and training rules