EXPERT SYSTEMS Part I. 2 What is there in an expert system? Components of an expert system Components of an expert system Knowledge acquisition and knowledge.

EXPERT SYSTEMSEXPERT SYSTEMSPart IPart I

2

What is there in an expert What is there in an expert system?system?

Components of an expert system Components of an expert system Knowledge acquisition and Knowledge acquisition and

knowledge engineering knowledge engineering Explanations Explanations User interface User interface Expert system shells Expert system shells Expert systems demosExpert systems demos

3

IntroductionIntroduction

Expert systems solve real problems which normally would require a specialized human expert (such as a doctor or a lawyer).

Knowledge characteristics:

• heuristic in nature, • based on ``rules of thumb''

4

DomainsDomains

• Medicine

• Engineering• Geology• Computer Science• Business• Law• etc

5

Types of problemsTypes of problems

diagnosis (e.g., of a system diagnosis (e.g., of a system fault, disease or student error)fault, disease or student error)

design (of a computer systems, design (of a computer systems, hotel etc)hotel etc)

interpretation (of, for example, interpretation (of, for example, geological datageological data

6

Rule Based Production Rule Based Production SystemsSystems

Most ES are Most ES are rule-based production systemsrule-based production systems

Their knowledge is represented in the form Their knowledge is represented in the form of of production rulesproduction rules to provide diagnoses or to provide diagnoses or advice on the basis of input data.advice on the basis of input data.

Production rules are Production rules are if-then rulesif-then rules

IFIF it rains, it rains, thenthen you will get wet you will get wet

IF IF it rains, it rains, thenthen you should wear a coat you should wear a coat

7

Components of an expert Components of an expert systemsystem

User

User

Inter-face

ExplanationsystemInference engine

Knowledge base editor

Case specific data: Working storage

Knowledge base

Expert system shell

8

Expert system knowledge Expert system knowledge basebase

Domain inference rules and object Domain inference rules and object descriptionsdescriptions

Case specific dataCase specific data Knowledge base editor: Knowledge base editor: creating creating

and updating the knowledge baseand updating the knowledge base

9

Knowledge acquisition Knowledge acquisition and knowledge and knowledge

engineeringengineering Domain expertDomain expert Knowledge engineerKnowledge engineer

Knowledge acquisitionKnowledge acquisition : : The The

process of "extracting" the knowledge process of "extracting" the knowledge

from the domain expert and representing from the domain expert and representing

the knowledge in a suitable form that the knowledge in a suitable form that

can be used by the expert systemcan be used by the expert system

10

11

Inference EngineInference Engine Forward chaining:Forward chaining: derive all conclusions derive all conclusions

given certain factsgiven certain facts Select applicable rulesSelect applicable rules Choose a rule (conflict resolution)Choose a rule (conflict resolution) Execute the rule and update the case-Execute the rule and update the case-

specific dataspecific data Backward chaining:Backward chaining: prove a given goal prove a given goal

by verifying the conditions in the IF-THEN by verifying the conditions in the IF-THEN rulesrules

Coping with uncertaintyCoping with uncertainty Mixed reasoningMixed reasoning

12

Forward ChainingForward Chaining

Forward chaining is a reasoning model Forward chaining is a reasoning model that works that works from a set of facts and from a set of facts and rules towards a set of conclusionsrules towards a set of conclusions, , diagnoses or recommendations.diagnoses or recommendations.

When a fact matches the antecedent When a fact matches the antecedent of a rule, of a rule, the rule firesthe rule fires, and the , and the conclusion of the rule is added to the conclusion of the rule is added to the database of facts.database of facts.

13

Conflict ResolutionConflict Resolution

More than one rule will may be usedMore than one rule will may be used A conflict resolution strategy decides A conflict resolution strategy decides

which rule to use.which rule to use. Strategies:Strategies:

give rules give rules prioritiespriorities and use the conclusion and use the conclusion that has the highest priority.that has the highest priority.

apply the rule with the apply the rule with the longest antecedentlongest antecedent apply the rule whose condition was apply the rule whose condition was most most

recently addedrecently added to the database. to the database.

14

Rete Algorithm for Forward Rete Algorithm for Forward ChainingChaining

Reorganize the rules in a network so that Reorganize the rules in a network so that the the conditions are linked to the rules, conditions are linked to the rules, i.e. rules are indexed by their i.e. rules are indexed by their antecedentsantecedents

When a new conclusion is made or a new When a new conclusion is made or a new fact arrives, examine only those rules fact arrives, examine only those rules whose antecedents contain the new whose antecedents contain the new conclusions or factsconclusions or facts

The algorithm makes forward chaining The algorithm makes forward chaining much faster at the expense of greater much faster at the expense of greater memory requirementsmemory requirements

15

Backward ChainingBackward Chaining

Used when a particular conclusion is Used when a particular conclusion is to be proved.to be proved.

Works back from a conclusion Works back from a conclusion towards the original facts.towards the original facts.

When a conclusion matches the When a conclusion matches the conclusion of a rule in the database, conclusion of a rule in the database, the system tries to prove the the system tries to prove the antecedents of the rule.antecedents of the rule.

16

Coping with uncertaintyCoping with uncertainty

The ability of the system to The ability of the system to reason with rules and data which reason with rules and data which are not precisely knownare not precisely known. .

Certainty factorsCertainty factors:: Apply to the known facts: how Apply to the known facts: how

much we trust the recorded factsmuch we trust the recorded facts Apply to the rules: how much we Apply to the rules: how much we

trust the rulestrust the rules

17

Meta RulesMeta Rules Meta rules : knowledge about how the Meta rules : knowledge about how the

system works.system works.

Example of meta rules:Example of meta rules:

The rules that determine the conflict The rules that determine the conflict resolution strategy resolution strategy

Meta rules might define that knowledge Meta rules might define that knowledge from Expert A is to be trusted more than from Expert A is to be trusted more than knowledge from Expert B.knowledge from Expert B.

18

Meta RulesMeta Rules

Implementation of meta rules:Implementation of meta rules: Implicit – in the inference Implicit – in the inference

engine (e.g. conflict resolution)engine (e.g. conflict resolution)

Explicit - treated by the system Explicit - treated by the system like normal rules, but are given like normal rules, but are given higher priorityhigher priority

19

Mixed ReasoningMixed Reasoning

Collect initial data and use Collect initial data and use forward reasoning to determine forward reasoning to determine a set of possible conclusionsa set of possible conclusions

Use backward reasoning to prove Use backward reasoning to prove or disprove each conclusionor disprove each conclusion

20

Example of Mixed Example of Mixed Reasoning in Medical ESReasoning in Medical ES

Collect data to build hypothesesCollect data to build hypotheses For each hypothesis H:For each hypothesis H: If H is in the facts database, it is proved. If H is in the facts database, it is proved. Otherwise, if H can be determined by asking a Otherwise, if H can be determined by asking a

question, then enter the user’s answer in the question, then enter the user’s answer in the facts database. Hence, it can be determined facts database. Hence, it can be determined whether H is true or false.whether H is true or false.

Otherwise, find a rule whose conclusion is H. Now Otherwise, find a rule whose conclusion is H. Now apply this algorithm to try to prove this rule’s apply this algorithm to try to prove this rule’s antecedents.antecedents.

If none of the above applies, we have failed to If none of the above applies, we have failed to prove H.prove H.

21

ExplanationsExplanations

Why Why a particular question is askeda particular question is asked How How a conclusion has been reacheda conclusion has been reached Why notWhy not some other conclusion some other conclusion What ifWhat if a particular symptom is present a particular symptom is present TerminologyTerminology – explain the terms used by – explain the terms used by

the systemthe system

22

User interfaceUser interface End-usersEnd-users Knowledge engineerKnowledge engineer ExpertsExperts

Requirements:Requirements: Self-consistentSelf-consistent Consistent with widely used interface Consistent with widely used interface

systemssystems Provide navigation helpProvide navigation help Incorporate user modelIncorporate user model

23

Expert System ShellsExpert System Shells

The shell is a piece of software which The shell is a piece of software which containscontains

the user interface, the user interface, a format for declarative knowledge in the a format for declarative knowledge in the

knowledge base, and knowledge base, and an inference engine.an inference engine.

The knowledge engineer uses the shell to The knowledge engineer uses the shell to

build a system for a particular problem build a system for a particular problem

domain.domain.

24

CLIPSCLIPS

CLIPS is C Language CLIPS is C Language Integrated Production Integrated Production System – System – an expert system shell.an expert system shell.

CLIPS uses a CLIPS uses a LISP-like notation LISP-like notation to enter rules.to enter rules.

25

CYCCYC

A frame based production system.A frame based production system.

Uses a database of over 1,000,000 Uses a database of over 1,000,000 facts and rules, encompassing all facts and rules, encompassing all fields of human knowledge.fields of human knowledge.

CYC can answer questions about all CYC can answer questions about all kinds of knowledge in its database, kinds of knowledge in its database, and can even understand analogies, and can even understand analogies, and other complex relations.and other complex relations.