Knowledge Representation zRepresentational adequacy ydeclarative, procedural zInferential adequacy ymanipulate knowledge yincorporate new knowledge.

Knowledge Representation

Representational adequacy declarative, procedural

Inferential adequacy manipulate knowledge incorporate new knowledge

Types of Knowledge

Simple factsComplex organized knowledgeprocedure - how to knowledgemeta-knowledge

Semantic Data ModelsHigh level model of model of

conceptual modelNot tied to implementation concernsFocus on

expressiveness simplicity concise formality

Semantic Nets

Nodes represent ObjectsLinks or Arcs represent Relationships

“instance of” - set membership “is a” - inheritance “ has a” - attribute descriptors “part of” - aggregation

Has a

Part-of

Instance of

Is a

Semantic NetsAdvantages Disadvantages

Flexibleeasy to understandsupport inheritance“natural” way to

represent knowledge

Hard to deal with exceptions

procedural knowledge difficult to represent

no standards for defining nodes or relationships

Classes, Objects, Attributes, Values - Object Orientation

Classes describe common properties of objects

Objects may be physical or conceptual

Attributes are characteristics of objects

Values are specific measures of Attributes for specific instances

Classes

Specify common properties of instancessupport hierarchical classificationsuperclass / subclass

subclass may be more refined version each subclass inherits operations and

attributes of its ancestors subclass may have its own operations and

attributes

Objects or Instances

Refers to things identified in model of conceptual model may be tangible (equipment, part,

orders, squashed bananas) may be mental constructs

Class vs instances

instances

Inheritance

Sharing attributes and behaviors within a class of objects

Person

customer

Employee

SalesPerson Manager

Sale Manager

Encapsulation

Attributes and behaviors (methods) integrated with the classes and objects

Attributes:size, location, appearance

Polymorphism

Each object responds in its unique way to messagesWhen changed method

When needed method

Object-Orientation

Tool for managing complexityemphasis on object structurespecify “what is”mapped directly from semantic net

Rule Representations

Rules are called productionsRule have two parts

condition part, premise -> IF action part ,conclusion-> THEN

The action can add a fact to the knowledge base, start a procedure or display a screen

Rules represent knowledge

Apply O-A-V framework (object-attribute-value)

IF air vehicle is a plane AND plane maximum altitude is 40000 AND plane manufacturer is Boeing THEN ASK Flight Display 15

Representing knowledge

Abstracting with rules translate quantitative to qualitative define technical terms support generalized reasoning

make rules for user easy to understand help user follow decision logic

Rule for understanding

Quantitative to Qualitative qualitative language is easier to

understand interpretation of numerical data make user feel comfortable with decision

logicIf temperature > 200 and humidity is

85% then machine is slightly overheated

Definitional Rules

Help communicate and train usersHelp user understand vocabulary Promotes common agreement on

terms for expert, user and knowledge engineer

IF you want more than one source file of classes THEN use package keyword

Rules support Generalizations

Allow reasoning with from specialization to generalizations

Support classification of objects at higher levels

Support refinements

If pump operation temperature is over 300AND water mixture pH > 5.2THEN replace pump bearing and oil

Surface KnowledgeSurface Knowledge

•Hard to understand•Difficult to learn reasoning strategies•hard to update and expand knowledge base

Hierarchical Classification

Feature abstractions Solution abstractions

Features Recommendations

generalize

Heuristic Match

refine

Abstraction draws out important aspects

Deep knowledge

Hot Pump Low Temp

Poor Oil Viscosity

Lubrication defect

causescauses

Is a

water mixture pH > 5.2temperature is over 300

Reasoning at higher level

Lubrication defectrequires

Maintenance

Fix heatdamage

Replace bearingand oil

Type of

Remedy

Modular style - easy to add, update and delete

natural for many problem domains

uncertain knowledge may be represented

May be difficult to understand

may demonstrate unpredictable behavior

extra effort required to representing structural knowledge

Rules Advantages Disadvantages

Programming by descriptiondescribe the problem’s factsbuilt in inference engine combines

and uses facts and rules to make inferences

Predicate Logic

Prolog Programming

Declaring facts about objects and their relationships -> likes (john,mary)

Defining rules about objects and relationships

Asking Questions about objects

sister-of(X,Y) :- female(X), parents(X,M,F), parent(Y,M,F)

Frames

Similar to objectshelps organize entitiespackages operations (demons)easy to modifyextensible through inheritance

Mammal Frame

Slot Values Default Demons

Skin Fur

Birth Live

Legs 4

Weight Computedemon

Frame - natural representation

Can accommodate a taxonomy of knowledge

contains defaults expectationsrepresent procedural and declarative

knowledge

Facets Inference Value Prompt

Exhaustive Conf

SearchOrder

Default Expand

WhenChanged

Init QueryFrom

WhenNeeded

Reinit Unknown

Facets - properties of slots