Introduction to ES Building an ES

1
Introduction to ES Building an ES
2

• AI Shells (ES Shells) such as EXSYS include
• all components of ES except knowledge base
• The major benefits of using ES shell
• Build specific ES rapidly and inexpensively.
• Permit non-programmers to build expert systems.

3

• The components of an ES

• Knowledge base
• Inference Engine
• Explnation Facilty
• User Interface
• Knowledge acquisition systems

4
Experts
Knowledge Acquisition System
Knowledge Base
User Interface
Inference Engine
User
Explanation System
Expert System Components
5
Knowledge Base Representing Knowledge with Rules

• In a rule-based ES, a rule is a conditional
  statement of IF.. THEN
• IF Socrates is a man
  
  THEN
  Socrates is mortal
• The ?F?side of a rule is also called the
  antecedent. The THEN side of a rule is called
  the consequent.
• Two Types of Rules (A Knowledge Rule and An
  Inference Rule)
• A Knowledge Rule A Rule Stored in the Knowledge
  Base (IF you study hard and you are lucky THEN
  you get an A in MG375).
• An Inference Rule A rule about rules (Meta
  Rule) (IF you cannot verify the antecedent , ask
  the user)

6

• Two Types of Knowledge
• Knowledge is a collection of specialized facts,
  procedures, and judgmental rules.
• Declarative (descriptive) knowledge
• description of the state of objects, facts (more
  commonly referred to data or information).(e.g.,
  IF the day is sunny, THEN the temerature will be
  high)
• declarative knowledge is stored in the knowledge
  base
• Procedural knowledge (step-by-step procedures,
  how-to-do instructions, knowledge to make a
  nuclear bomb).

7
Inference Engine and Reasoning with Logic

• The Inference Engine is the ES component
  responsible for drawing conclusions through
  processing the rules in the knowledge base.
• There are two basic reasoning methods.
• Deductive reasoning moves from a general
  principle to a specific inference. e.g., -- It
  does not snow if the temperature is above 32
  degrees F. It is 40 degrees F. Therefore, there
  is no snow.
• Inductive reasoning uses some established facts
  to draw general conclusions..
• For performing either deductive or inductive
  reasoning, several resoning procedures allow the
  manipulation of logical expression.

8
Inference Engine

• The Inference Engine is the ES component
  responsible for drawing conclusions through
  processing the rules in the knowledge base.
• Conclusion are reached by the application of two
  stanadard rules.
• Modus Ponens is a basic reasoning procedure --
  In a rule based ES, If A is true, THEN B is true,
  and vice versa.
• IF A, THEN B. A true, Therefore Btrue.

9

• Hypothetical Syllogism is other rule of
  inference.
• IF A, THEN B
• IF B, THEN C
• Therefore IF A, THEN C
• e.g., IF interest rates rise, THEN the value of
  the dollar increases.
• IF the value of the dollar increases, THEN U.S.
  export quantitires decrease.

10
AND/OR Truth Tables

• A conditional statement having more than two
  propositions (statements that can be assigned a
  value of .T. or .F.) employs the connectives AND
  and OR to link them.
• AND Truth Table
• OR Truth Table

11

• Problem solving strategies

ES employs two strategies in solving hypothetical
syllogisms Backward Chaining and forward
chaining. Data lt------ backward reasoning
lt-------------- a Goal Data
(the first rule) --gt forward reasoning
(chaining) ---gta Goal LEVEL5 supports both
strategies. Default is backward reasoning.
12
Developing an ES (a Knowledge base)- An Example

• TITLE Loan Advisor
• !
• ! Goal Statement
• 1. Approve the loan request
• 2. Deny the loan request
• !

13
! production rules-----------

• !
• RULE 1 Loan Decision
• IF Applicant's Job is stable
• OR applicant satisfies networth requirements
• AND Residency requirements have been met
• AND Debt service levels are acceptable
• AND Credit Report is satisfactory
• AND Prospects for repayment of the loan are good
• THEN Approve the loan request

14

• RULE 2 Job Stability
• IF Years at present job gt 2
• OR Years at previous job gt 2
• THEN Applicant's Job is stable
• RULE 3 Residency Requirements
• IF Years the applicant has lived in area gt 2
• THEN Residency requirements have been met

15

• RULE 4 Deny Loan Request
• IF NOT Approve the loan request
• THEN Deny the loan request
• !
• END

16
Structure (Components) of expert systems.

• User
• User Interface is a software that facilitates
  friendly, easy-to-use communication between the
  user and the computer via a natural language,
  menus, and graphics.
• Knowledge Acquisition Subsystem is to create,
  add, edit (change) the knowledge base.
• Knowledge Base contains Facts (What is known
  about the Domain Area) and Rules that directs the
  use of knowledge to solve specific problems.
• such as the following
• Why was a certain question asked by the expert
  systems?
  
  
  
  How was a certain conclusion reached?

17
Inference Engine is the ?rain?of the ES. It is
also known as the rule interpreter in rule-based
ES. Inference engine is a computer program that
provides a methodology for reasoning about
information in the knowledge base. Explanation
Subsystem (Justifier) helps the user by
explaining the basis for recommendation that are
produced and by interactively answering questions
18
What is the Size of a knowledge base? (Wang)

• San Wa Bank? Personal FPS (selecting investment
  instrumnt) 130 rules
• AX Authorizer? Assistant 1500 rules
• Airconditioning Diagnostics ES 500 Rules
• CCH-ES 800 Rues
• Comprehensive PFPS (pension, retirement, estate
  planning, etc.) Several Thosnds
  rules
• Strategy Consulting ES 2000 rules in 44 KB
• Cokpit Crew Scheduling Systems 45000 frames,
  7000 rules

19
(No Transcript)
20

• The systems knowledge base is
• R1 IF candidates technical qualifications are
  excellent or average, THEN candidate is a
  technically acceptable prospect
• R2 IF candidates technical qualifications are
  minimal, THEN candidate is not a technically
  acceptable prospect
• R3 IF candidates external interests are diverse,
  non-business related, or multiple, strictly
  business related subjects, THEN candidate is
  rounded acceptably (interests), ELSE reject
  probability 7/10
• R4 IF candidate is mentally quick, THEN candidate
  is mentally agile
• R5 IF candidate appears mentally slow, THEN
  return for more detailed interview probability
  4/10 and reject probability 6/10

21

• R6 IF candidates verbal skills are excellent or
  average, THEN candidate is verbally qualified
• R7 IF candidates interpersonal skills are
  excellent or average, THEN candidate is
  interpersonally acceptable
• R8 IF candidates technical qualifications are
  excellent and candidate is rounded acceptably
  (interests) and candidate is mentally agile and
  candidate is interpersonally acceptable and
  candidates verbal skills are excellent, THEN
  unconditional hire, probability 8/10, ELSE
  unconditional hire probability 0/10
• R9 IF candidate is NOT a technically acceptable
  prospect and candidate is NOT verbally qualified
  or NOT rounded acceptably (interests) or NOT
  mentally agile or NOT personally acceptable, THEN
  reject probability 10/10, ELSE reject
  probability 3/10

22

• R10 IF candidate is a technically acceptable
  prospect and candidate is verbally qualified and
  candidate is rounded acceptably (interests) and
  candidate is mentally agile and candidate is
  interpersonally acceptable, THEN hire, but at
  reduced rate probability 7/10 ELSE, return for
  more detailed interview probability 6/10.
• R11 IF candidate is not a technically acceptable
  prospect and candidate is rounded acceptably
  (interests) and candidate appears mentally agile
  and candidates verbal skills are excellent, THEN
  return for more detailed interview probability
  8/10

23
Hire at reduced rate

• excellent
• average

Technically acceptable prospect
Technical qualification
OR
AND

• excellent
• average

Verbal Skills
OR
Verbally qualified
Hire at reduced rate

• divers, non-business related
• multiple, strictly business related

OR
Interests
Rounded acceptably
Mental quickness
Mentally agile

• quick

• excellent
• average

OR
Interpersonal Skills
Interpersonally acceptable
24
Return for more detailed interview
Technical qualifications

• minimal

• divers, non-business related
• multiple, strictly business related

AND
OR
Interests
Return for more detailed interview
Rounded acceptably
Mental quickness
mentally agile

• quick

Verbal skills

• excellent

25
Unconditional Hire
By Tim Richter
Technical Qualifications
Technically Acceptable Prospect

• Excellent

AND

• Diverse, non-business related
• Multiple, strictly business related

OR
Interests
Rounded Acceptably
Unconditional Hire
Mental Quickness

• Quick

Mentally Agile
OR
Interpersonal Skills

• Excellent
• Average

Interpersonally Acceptable
Verbal Skills

• Excellent

26
Is the battery good?
Diagnosing a Car Starting Problem
Yes
No
Does the engine crank?
Replace the battery (1)
No
Yes
Is there a clicking sound?
Is the gas tank empty?
Yes
No
No
Yes
Are cables clean?
Clean cables (4)
Does the engine sputter?
Add fuel (5)
No
Yes
Yes
No
Check filter and fuel pump (6)
Check ignition switch (3)
Clean cables (2)
Is the filter clean?
Yes
No
Does the car smoke?
Replace filter (7)
Yes
No
Check coil and wiring (9)
Adjust the choke (8)