KNOWLEDGE

ACQUISITION &
ANALYSIS
1
OVERVIEW

“An expert system gains its power

from the knowledge it contains”

Durkin, (1994)

2
INTRODUCTION
Knowledge – a collection of facts, rules, and concepts
used to reason with.

Knowledge acquisition
 A generic term – how to transfer of knowledge is achieved.
 the process of acquiring, organizing, and studying knowledge.
 the transfer and transformation of potential problem solving expertise from some
knowledge source to a program.

3
TYPES OF KNOWLEDGE
Declarative - knowing that something is true or false Example: Don’t put your
fingers in a pot of boiling water.
Procedural - knowing how to do something . Example: knowing how to boil a
pot of water.
Meta knowledge - Knowledge about the other types of knowledge and how to
use them. Example: if you want to know about heart attack, please read this
book
Heuristic - Rules of thumb that guide the reasoning process. Example: the
clouds looks dark and heavy, … heavy rain might fall…
Structural – Describe an expert overall mental model of the problem. Example:
in hierarchy, from concept to subconcept and object and they are related.
Tacit - knowledge is a something that cannot be expressed by language.
Example: knowing how to move your hand.

4
KNOWLEDGE
ELICITATION
Knowledge acquisition is the process of acquiring knowledge from
the expert.
Knowledge elicitation is the process of extracting and formalizing
the knowledge of a domain expert for use by an AI system.

5
KNOWLEDGE
ELICITATION
Knowledge Elicitation – the knowledge transfer is
accomplished by a series of interviews between a domain
expert and a knowledge engineer who then writes a computer
representing the knowledge.

Required close collaboration between the domain expert and

knowledge engineer.

 Getting information from expert in a particular area, then put the

information into a computer.

 Also involves computer support.

6
KNOWLEDGE
ELICITATION
Knowledge acquisition is a difficult task.

“Knowledge acquisition is a bottleneck in the construction of

expert systems. The knowledge engineer's job is to act as a
go-between to help build an expert system. Since the
knowledge engineer has far less knowledge of the domain
than the expert, however, communication problems impede
the process of transferring expertise into the program.”

(Hayes-Roth et al., 1983)

7
KNOWLEDGE
ELICITATION - SOURCES
Source of knowledge:

Report

Books

Domain Expert
Database
Report

8
KNOWLEDGE
ELICITATION - SOURCES
Expert End-user
• The primary source of knowledge. • Important additional source of information.
• Need to capture the expert’s unique • View the problem from high level –
expertise. considering the major issues.
• View a problem from a low level – • Important when we need an initial general
considering only the important details. understanding of the problem.

Multiple experts Literature

• Sometimes more than one experts is • Additional source of information –
required to collect specialized knowledge on documents such as reports, regulations,
some sub-problem or to verify the guidelines and books.
knowledge collected from the single expert. • Also help to define and clarify the
terminology of the domain.
• Provide insight into the major issues you
will need to address, and details on the
knowledge that will eventually be used in
the final system. 9
KNOWLEDGE
ELICITATION TASKS
Knowledge elicitation tasks
Collect
 Is the task of acquiring knowledge from the expert.
 Interpret
 Involves a review of the collected information and
the identification of key pieces of knowledge.
 Analyze
 Identify the important concepts used by the expert.
 Can determine concept relationships and how the
expert uses them to solve the problem.

10
KNOWLEDGE
ELICITATION TASKS
Knowledge elicitation tasks
 Design
 Previous activities should give us some new
understanding of the problem that can aid further
investigations.
 Provides guidance in designing new techniques for
collecting additional knowledge.

11
 DIFFICULTIES
One of the problems is the difficulty that experts have in
describing exactly

HOW they do what they do,

especially with respect to their use of

1) judgment,
2) experience, and
3) intuition. 12
DIFFICULTIES
Unaware of knowledge - An expert often compiles the problem-
solving knowledge into a compact form – often make mental leaps
over important issues.
Unable to verbalize knowledge - Difficult to verbalize the tasks.
Provides irrelevant knowledge.
Provides incomplete knowledge - Omission or unaware of knowledge
used.
Provides incorrect knowledge - Expert is uninformed or mistake
during introspection.
Provides inconsistent knowledge - Inconsistent with earlier statements.

13
DIFFICULTIES - COMMON
PROBLEMS
Problem in KA Methods/Solution
Incompatible data Data transport tools
Inconsistent data Eliminate different
representation
Missing data Check blank fields

A cheese factory had an experienced cheese-tester who was

approaching retirement age. The factory manager decided to replace
him with an “intelligent machine”. The human tester tested the cheese
by sticking his finger into a sample and deciding if it “felt right”. So it
was assumed the machine had to do the same – test for the right
surface tension. But the machine was useless. Eventually, it turned
out that the human tester subconsciously relied on the cheese’s smell
rather than on its surface tension and used his finger just to break the
crust and let the aroma out.
14
KNOWLEDGE ELICITATION
MEMBER ROLES

• Provide problem
• Provide
Knowledge overview
knowledge
Engineer • Help define
interface
• Aid • Collect knowledge requirements
interpretati • Interpret and • Define explanation
on/ analysis analyze knowledge requirements
• Aid testing • Design new session • Help define in-
Domain • Manage project place
• operational
Expert • Maintain requirements
corporative effort End User
• Aid system
development

Roles of elicitation team members

15
METHOD
Introspection – Examining the thought processes the expert builds a
system which he believes effectively replicates the thinking
processes
Interviewing –The knowledge engineer extract the knowledge
provided during interview session and build the system in a manner
which he believes is similar to the way expert thinks.
Observation – The expert closely observed whilst at work
Induction – The process of converting a set of examples into rules.
Protocol Analysis – Expert is asked to perform task and to verbalize
his thought process
Prototyping – An extension of the interviewing technique. Here the
expert works with the knowledge engineer in building a system
16
METHOD-INTERVIEW

Structured Interview Unstructured Interview

• Maintains a focus on one issue • To allow the expert to discuss
at a time. a topic in a natural manner.
• Elicits specific details on • Can get a conceptual
given issue before moving on understanding of the problem
to other points. and insight into the general
• Focus on single issue – get problem-solving strategies
detail knowledge – uncovers used by the expert.
related topics that can be • Ask a general question about
explored some broad problem issue –
using a prompt or indirect
question.

17
METHOD-INTERVIEW
Conducting interview
 Guidelines for beginning interview

 Establish a comfortable setting by discussing a topic of

personal interest
 Avoid an initial subject related to the project
 Review the session’s objective and agenda
 Resolve any misunderstandings and solicit suggestions

18
 METHOD-INTERVIEW:
Type QUESTIONS
Purpose COMMONLY USEDForm
Direct Obtain specific information on some What does… mean?
known issue. Is … true?
Questions commonly used in an interview
What is the value of…?
Indirect Obtain general information on What issues are considered
concepts and problem solving for …?
strategies. How do you determine …?
What do you look for when
…?
Probes Probe deeper into an established Can you explain …?
issue Can you discuss …?
Prompts Direct interview into a new area Can you discuss …?
Can you return to …?

19
 METHOD-INTERVIEW
Guidelines for forming a Good Question
 Avoid multiple part questions.

“Why are the pressure and the temperature of the pump…?”

 Avoid leading questions – imply the expected answer and reflect your
possible bias on a given issue.

“Wouldn’t you check the blood pressure first?”

 Do not phrase questions in the negative – can confusing and cause

difficulties with interpretation that lead to misleading results.

“Is it false that the PC burn when the room temperature

is not high is not true?”

20
METHOD-INTERVIEW
Guidelines for forming a Good Question
 Avoid universal questions – presumes that the issues applies to every
situation, when it may only apply to some unique set of events.

“Are you always check your patient heart beat?”

 Avoid the use of terms foreign to the expert.

“You use forward or backward reasoning to solve the problems?”

 Watch for questions that use subjective words.

“If the toaster is really hot, Do you shut it fast?”

21
METHOD-INTERVIEW
Guidelines for ending an interview

 Summarize key findings

 Review any outstanding questions raised during the interview
and comment on how they will be answered.
 Review any outstanding agenda items and discuss how they
will be dealt with in the future.
 Review future tasks for each team member before next
meeting.
 Maintain open communication lines for aiding future tasks.
 Extend appreciation to members for their time and effort.

22
METHOD-INTERVIEW

Problems with interviewing

 Recalling procedural knowledge
 Ineffective long-term memory
 Verbalizing manual tasks
 Verbalizing compiled knowledge
 Lacks context

23
METHOD-CASE STUDY

Case study
 Observational
 Asking the expert to solve the problem - while you observe.
 Retrospective
 Expert is asked to review the case and explain in retrospect how
the problem was solved.

24
METHOD COMPARISON
General Observations in comparing elicitation
 Case studies are better than interviewing techniques or
obtaining procedural knowledge.
 Structured interviews are the best choice for uncovering
concepts, objects, and relationships.
 Unstructured interviews are best for obtaining insight
into general concepts and problem solving methods.
 The observational case study method is, in general,
better than the retrospective method for obtaining
problem details and avoiding knowledge elicitation
difficulties.
 Familiar case studies are best for obtaining common
domain concepts and typical problem solving methods.
27
METHOD COMPARISON

General Observations in comparing elicitation

 Unfamiliar case studies are best for uncovering basic
problem principles.
 Unstructured techniques are poor for avoiding
elicitation problems
 In general, case studies are better than interviewing
techniques for avoiding elicitation difficulties.
 Unfamiliar case studies are best for avoiding elicitation
problems.

28
KNOWLEDGE ANALYSIS

Must interpret and analyze the collected

information – to form theories on their
organization.
Producing transcript
 Should produce a complete and exact transcript of the recorded
session:
 To avoid lose valuable information.
 To uncovers subjective terms that the expert might use to qualify an
issue – important if inexact reasoning is used.

 Producing transcript is time consuming.

29
KNOWLEDGE ANALYSIS

Interpreting the transcript

 Reviewing the transcript and identifying the key pieces of
knowledge.
 Identify any issues that need further clarification by the expert.
 Mark and highlight the knowledge in the transcript and label with
the type of knowledge it represents (e.g: concept,
strategy, etc.).

30
 KNOWLEDGE ANALYSIS
Analyzing the transcript
 Question:
“How does this new information fit in with what is already known about
the problem?”

 Answer:
 Must address how this new information supports or relates to previously
known information.
 Involves determining
 How this new information is naturally linked to already known related concepts,
 How it adds details to established information, or
 How it uncovers new concepts or rules.

 Ask the expert to review the result of this analysis to confirm the findings
and make adjustments.
31
KNOWLEDGE ANALYSIS

Analyzing the transcript

 Knowledge analysis involve several steps:
Recording the knowledge
Relating the knowledge
Reviewing the knowledge
Structuring the knowledge graphically

32
KNOWLEDGE ANALYSIS
Analyzing the transcript
 Recording the knowledge

 Record each new pieces of information with other similar

pieces discovered during earlier interviews – in
knowledge dictionary.

 Knowledge dictionary is a part of the project

documentation that has sections for each type of
knowledge, e.g. concepts, objects, rules, etc.- should also
have reference to its source.
33
KNOWLEDGE ANALYSIS
Analyzing the transcript
 Relating the knowledge

 Sorting and relating the similar pieces of information

collected with previously recorded information.

 Uncovers the natural structure and organization of the

collected pieces of knowledge.

34
KNOWLEDGE ANALYSIS
Analyzing the transcript
 Reviewing the knowledge

 Review what have been collected and what are

need to be collected.

 Highlight the areas that need further pursuit.

35
KNOWLEDGE ANALYSIS
Analyzing the transcript
Structuring the knowledge graphically

 Relate the information collected in a graphical fashion.

 Provide visual perspectives of the important knowledge and its

organization.

 Approaches:
 Cognitive maps
 Inference networks
 Flowcharts
 Decision trees

36
KNOWLEDGE ANALYSIS
Analyzing the transcript
Cognitive Maps

 One of the most common methods used.

 Show a natural relationships between concepts of
objects.
 Composed of nodes and arcs that link related nodes.
 The nodes can represent either abstract (e.g: birds,
computers, car, conceptual understanding, etc.) or
concrete objects (e.g.: specific bird).
37
KNOWLEDGE ANALYSIS
Analyzing the transcript
Cognitive maps of employee

38
KNOWLEDGE ANALYSIS
Analyzing the transcript
Inference Networks

 Provide a graphical representation of the system’s

rules, with the antecedents and consequences of the
rules drawn as nodes and their supporting relationships
drawn as links.

39
 KNOWLEDGE ANALYSIS
Analyzing the transcript
 Inference Networks - Example a set of rules for rain prediction

Rule 7
IF Barometric pressure is falling
AND Wind conditions indicate rain
AND Temperature is moderate
THEN Weather prediction is rain

Rule 8
IF Wind is gusty
OR Wind direction is from the east
THENWind conditions is indicate rain

Rule 9
IF Wind speed is > 5 knots
THENWind is gusty

Rule 10
IF Temperature is between 60 and 80 degrees
THENTemperature is moderate

40
KNOWLEDGE ANALYSIS
Analyzing the transcript
 Inference Networks - Example for rain prediction
Rule 7
Weather
prediction
is rain

Rule 8
AND Rule 10
Barometric Wind Temperature
pressure is conditions is moderate
falling indicate
rain

Rule 9
OR
Wind is Wind 60<Temp<80
gusty direction is
from the
east
41
Wind speed > 5 knots
 EXERCISE
Represent the rules below using inference network diagram

42
KNOWLEDGE ANALYSIS
Analyzing the transcript
Flowcharts
 Present a sequence of steps that will be performed.
 Standard technique used by conventional programmers
– to show the sequence of operations of the program.
 Links between the blocks show the natural order of the
execution of the blocks or branches to other points
within the program.

43
KNOWLEDGE ANALYSIS
Blood Disease
Diagnosis
Analyzing the transcript
 Flowcharts
Obtain Initial
Data

Consult NO Can Form

Specialist Hypothesis

YES
Ask Questions
Related to
hypothesis

44
KNOWLEDGE ANALYSIS
Analyzing the transcript
Decision Trees

 Graphical representation of a problem’s search space.

 Is composed of nodes and arcs linking related nodes.
 Each nodes represents a decision issue and the arcs
represent possible values for each issue.
 Can trace through the tree to arrive at a solution to the
problem.

45
KNOWLEDGE ANALYSIS
Analyzing the transcript
 Decision Trees - Decision tree for diagnosing problems
Terminals

Not Loose Loose

Battery Tighten
Voltage Terminals

> 12 < 12

Distributor Charger

Ok Bad Bad Ok

Check Replace Replace Replace

Starter Distributor Charger Battery 46