Expert System

Expert Systems (ES)

● Expert systems are knowledge based programs which
provide expert quality solutions to the problems in
specific domain of applications.
● The core components of expert system are
− knowledge base and
− navigational capability (inference engine)
● Generally its knowledge is extracted from human
experts in the domain of application by knowledge
Engineer.
− Often based on useful thumb rules and experience rather
than absolute certainties.
● A process of gathering knowledge from domain
expert and codifying it according to the formalism
is called knowledge engineering.
 Phases in building Expert System
● There are different interdependent and overlapping
phases in building an expert system as follows:
● Identification Phase:
− Knowledge engineer finds out important features of the
problem with the help of domain expert (human).
− He tries to determine the type and scope of the problem, the
kind of resources required, goal and objective of the ES.
● Conceptualization Phase:
− In this phase, knowledge engineer and domain expert decide
the concepts, relations and control mechanism needed to
describe a problem solving.
 Cont…
● Formalization Phase:
− It involves expressing the key concepts and relations in some
framework supported by ES building tools.
− Formalized knowledge consists of data structures, inference
rules, control strategies and languages for implementation.
● Implementation Phase:
− During this phase, formalized knowledge is converted to
working computer program initially called prototype of the
whole system.
● Testing Phase:
− It involves evaluating the performance and utility of prototype
systems and revising it if need be. Domain expert evaluates
the prototype system and his feedback help knowledge
engineer to revise it.
 Expert System Architecture
Expert System
Inference Engine
Special Interfaces
Inference & Control

Human Case History

Expert Knowledge Acquisition
& Learning Module
Knowledge Base

Static database
User User Interface

Dynamic database
(working memory)

Explanation Module

Fig: Architecture of Expert System

 Knowledge Base (KB)
· KB consists of knowledge about problem domain in the
form of static and dynamic databases.
· Static knowledge consists of
− rules and facts which is complied as a part of the system and
does not change during execution of the system.
· Dynamic knowledge consists of facts related to a
particular consultation of the system.
− At the beginning of the consultation, the dynamic knowledge
base often called working memory is empty.
− As a consultation progresses, dynamic knowledge base grows
and is used along with static knowledge in decision making.
· Working memory is deleted at the end of consultation
of the system.
 Inference Engine
· It consists of inference mechanism and control strategy.
· Inference means search through knowledge base and
derive new knowledge.
· It involve formal reasoning involving matching and
unification similar to the one performed by human
expert to solve problems in a specific area of
knowledge.
· Inference operates by using modus ponen rule.
· Control strategy determines the order in which rules are
applied.
· There are mainly two types of control mechanism viz.,
forward chaining and backward chaining.
Knowledge Acquisition
● Knowledge acquisition module allows system to
acquire knowledge about the problem domain.
● Sources of Knowledge for ES
− text books, reports, case studies,
− empirical data and
− domain expert experience.
● Updation of Knowledge can be done using
knowledge acquisition module of the system.
− insertion,
− deletion and
− updation of existing knowledge
Case History
● Case History stores the file created by inference
engine using the dynamic database created at the
time of consultation.
● Useful for learning module to enrich its knowledge
base.
● Different cases with solutions are stored in Case
Base system.
● These cases are used for solving problem using
Case Base Reasoning (CBR).
Explanation module
● Most expert systems have explanation facilities that
allow the user to ask the system why it asked some
question, and how it reached to conclusion.
● It contains 'How' and 'Why' modules attached to it.
− The sub-module ‘How’ tells the user about the process
through which system has reached to a particular solution
− ‘Why' sub-module tells that why is that particular solution
offered.
● It explains user about the reasoning behind any
particular problem solution.
● Questions are answered by referring to the system
goals, the rules being used, and any existing problem
data.
Explanation Modules Responses
User Why module
System Do you have running nose?
User Yes.
System I am trying to prove that you have cold. Do you have headache?
User Yes
System Do you have sore throat?
User No
System Then you do not have cold. Do you have fever?
User Yes
System Now I am looking at the possibility of measles. Do you have cough?
User Yes
System I can infer measles using rule “If symptoms are fever, cough,
running_nose, then patient has measles” measles is concluded.
User How Module
System Since you have fever, running_nose and cough and there is a rule “If
symptoms are fever, cough, running_nose, then patient has measles”.
So measles is concluded for you.
 User Interfaces
● Allows user to communicate with system in interactive
mode and helps system to create working knowledge
for the problem to be solved.
Dialogue Module (User Interface)

System Do you have fever?

User Yes
System Do you have bad throat?
User No
System Do you have cough?
User Yes
System Are you suffering from running nose?
User Yes
System Are you suffering from headache?
User No
Special interfaces
● It may be used for specialized activities such as
handling uncertainty in knowledge.
● This is a major area of expert systems research that
involves methods for reasoning with uncertain data
and uncertain knowledge.
● Knowledge is generally incomplete and uncertain.
● To deal with uncertain knowledge, a rule may have
associated with it a confidence factor or a weight.
● The set of methods for using uncertain knowledge in
combination with uncertain data in the reasoning
process is called reasoning with uncertainty.
Expert Systems Traditional Systems
1. Manipulates knowledge 1. Manipulates data
2. Problem-related expertise is stored in 2. Problem-related expertise is
data structures only, not in programs stored in both data structures and
(Example: Rules like ab are stored as programs
knowledge) (Example: Rules like ab are coded
as “if a then b” – making it a part of
the decision making logic in the
program)

3. Hence, if knowledge changes the 3. Hence, if knowledge changes,

knowledge base has to be updated the program has to be rebuilt
4. Allow use of confidences or certainty 4. Do not have confidences or
factors like humans. certainty factors like humans.
Example: If weather is humid, then it Example: If weather is humid, then
might probably rain. it might probably rain.
Weather(humid, today)rain(today) with Weather(humid, today)rain(today)
CF=0.9

5. Not necessarily 100% correct 5. 100% correct answers

answers
Characteristics of Expert Systems
1. Expertise
2. Symbolic Reasoning: Knowledge is
represented symbolically which can be
easily reformulated and reasoned
3. Self Knowledge: System should be able to
explain and examine its own reasoning
4. Learning Capability: Learns from mistakes
5. Ability to provide training: Explain reasoning
process
6. Predictive modelling power
Evaluation of Expert Systems
 Does the system make decisions the experts
agree to?
 Are the inference rules correct and complete
 Does the control strategy allow the system to
consider items in the natural order that the
expert prefers?
 Are relevant questions asked to the user in
proper order?
 Are explanation adequate for explaining how
and why concusions?
Evaluation of Expert Systems
Utility Evaluation
 Does the system help the user in some
significant way?
 Are the conclusions of the system organized
and ordered in a meaningful way?
 Is the system fast enough to satisfy the user?
 Is the user interface friendly enough?
Advantages of Expert Systems
 Scarce expertise made available.
 Reliability – provides consistent answers for repetitive
decsions, processes and tasks
 ES can work faster than people.
 Integration of expertise from different sources.
 Improved quality (good decision making, cost saving,
understanding and explaining logic and reasoning)
 Ability to work with incomplete information.
 Reduced system downtime (ES monitors or finds
faults).
 Training (users gain expertise from the ES).
 Makes expertise available in remote locations.
 Disadvantages of Expert Systems

 Not like human experts who are more creative and

flexible.
 Lacks common sense needed in decision making.
 May cause errors in the knowledge base and
decision making. Human experts are generally aware
of the scope and limitations of their knowledge.
 Cannot adapt to changing environment. When faced
with a new situation a human expert may develop a
new and brilliant approach to solving the problem.
Tools for Building Expert Systems
 Functional programming languages: LISP
 Logic programming languages: Prolog
 Sometimes object-oriented languages like C+
+, Java are used. Also C language is used
 ES shells: Contain built-in formalism for
knowledge representation and reasoning
mechanism. Example: EMYCIN - domain
independent version of the MYCIN
Truth Maintenance System (TMS)
● Truth maintenance system (TMS) works with inference engines
for solving problems within large search spaces.
● The TMS and inference engine both put together can solve
problems where algorithmic solutions do not exist.
● TMS maintains the beliefs for general problem solving systems.

Problem Solver

IE TMS

KB
TMS - Cont
● TMS can be used to implement monotonic or non-
monotonic systems.
● In monotonic system, once a fact or piece of
knowledge is stored in KB, it can not change.
− In monotonic reasoning, the world of axioms continually
increases in size and keeps on expending.
− Predicate logic is an example of monotonic form of
reasoning. It is a deductive reasoning system where new
facts are derived from the known facts.
● Non-monotonic system allows retraction of truths that
are present in the system whenever contradictions
arise.
− So number of axioms can both increase and decrease and
depending upon the changes in KB, it can be updated.
Example – Monotonic TMS
● Suppose we are given the premise set  = {P, W}
and the internal constraint set
{P  Q, (P  W)  R, (Q  R)  S}.
● TMS are able to derive S from these constraints and
the premise set .
● TMS should provide the justifications of deriving S
from constraints and premises.
● Therefore, for any given set of internal constraints
and premise set , if a formula S can be derived from
these, then justification functions generate a
justification tree for S.
Justification Tree

(Q  R)  S

{Q, R}

PQ Q R (P W)  R

{P} {P, W}
Non-Monotonic TMS
● TMS basically operates with two kinds of
objects
− ‘Propositions’ declaring different beliefs
and
− ‘Justifications’ related to individual
propositions for backing up the belief or
disbelief expressed by the proposition.

● For every TMS, there are two kinds of

justifications required namely ‘Support list’
and ‘Conditional proof’.
Support List (SL)
SL(IN-node)(OUT-node)
where
IN-node is a list of all IN-nodes (propositions) that
support the considered node as true. Here IN means
that the belief is true.
OUT-node is a list of all OUT nodes for the considered
node to be true. OUT means that belief is not true.
Node No. Facts/assertions Justification (justified belief)

1 It is sunny SL(3) (2,4)

2 It rains SL() ()

3 It is warm SL(1) (2)

4 It is night time SL() (1)

Conditional Proof
● A belief may be justified on the basis of several other
beliefs, by the conditional proof on one belief relative
to other beliefs, or by the lack of belief in some fact.
● These are justifications which support belief if a
specified belief follows from a set of other beliefs.
● Truth maintenance processing is required when new
justifications change previously existing beliefs.
● In such cases, the status of all beliefs depending on
the changed beliefs must be re determined.
● Dependency-directed backtracking is a powerful
technique based on the representations of the truth
maintenance system.
Conditional Proof
 CP-justifications take the form
(CP (consequent) (inhypotheses)

(outhypotheses))
 A CP-justification is valid if the consequent
node is in whenever
(a) each node of the inhypotheses is In and
(b) each node of the outhypotheses is Out.
Example: Consider scheduling a meeting, to
be held preferably at 10 A.M. in either room
813 or 801.
Dependency Directed Backtracking