Akuisisi Pengetahuan

(Knowledge Acquisition)
Departemen Ilmu Komputer IPB
2012

The heart of ES

Knowledge
Declarative vs Procedural Knowledge

Declarative Knowledge (knowledge of facts)

Procedural Knowledge (knowledge of how to do something)

Tacit vs Explicit Knowledge

Tacit (cannot be articulated easily)

Explicit (can be articulated easily)

Generic vs Specific Knowledge

Overview

Knowledge
Acquisition
What is it?
What are the
issues?
What are the
solutions?
What techniques are
used?

Knowledge Acquisition What is it?

Knowledge
People

Documents

Knowledge
Acquisition

Software

Knowledge Engineer

Knowledge
Repository

Tujuan Akuisisi Pengetahuan

Merumuskan pengetahuan (knowledge Base),

sehingga dapat diorganisasikan dalam

komputer
Mendapatkan pengetahuan, fakta aturan,
model dan cara pemecahan masalah dari
pakar

Akuisisi Pengetahuan
Mengumpulkan informasi dari berbagai

sumber/pakar untuk kemudian disimpan dalam

sistem komputer
Sumber pengetahuan: pakar/experts, buku,
technical reports, databases, forms dll.

Hayes-Roth et al (1983)

Knowledge Acquisition is a bottleneck in

the construction of ES. The KEs job is to act as
go-between to help build an expert system.
Since the knowledge has far less knowledge of
the domain than the expert, however,
communication problems impede the
process of transferring expertise into the
program

Knowledge Acquisition - Issues

Much vital knowledge is

Much
vital knowledge
is
in peoples
heads
in peoples
(experts)heads
(experts)
The most relevant and up-toKnowledge
date knowledge
The knowledge required forRepository
optimal task performance
What really happens (and
should happen)

Knowledge Acquisition - Issues

The repository should

contain
knowledge
The
repository
shouldthat
is..
contain knowledge
that
is..
Relevant, Vital, Best
Practice, Unambiguous
Stored in ways that are easy
to access and maintain
Presentable in ways that are
easy to understand

Knowledge
Repository

Knowledge Acquisition - Issues

Acquiring knowledge is
difficult, and can be very
time consuming, costly and
inefficient.

Knowledge
Repository
Need techniques that:

Take experts off the job for short

periods
Capture structured knowledge
efficiently
Focus on the essential knowledge
Can capture tacit knowledge

Tacit
TacitKnowledge:
Knowledge:

Deep
Deepknowledge
knowledgenot
not
consciously
consciouslyavailable
available
totothe
theexpert,
expert,hence
hence
difficult
difficultto
todescribe.
describe.

Penyebab sulitnya akuisisi pengetahuan

Expressing Knowledge
Transfer to a Machine
Number of Participants
Structuring Knowledge

Where to look for solutions?

Where to look for
solutions?

Experimental
Psychology

Artificial
Intelligence

Research in the 1980s found:

Techniques that capture tacit

knowledge
Different ways of representing
knowledge
Need range of techniques
Efficient acquisition methodologies

Knowledge
Repository

Knowledge Acquisition Solutions

Knowledge Models

Structured
representations
Clear and Unambiguous
Help elicit knowledge
Help validation
Help use and
maintenance

Efficient interviews
Acquire tacit
knowledge
Help experts recall
Help to focus the
expert
Generic tasks

Psychological
Techniques

Kemampuan yang harus dimililiki oleh KE

Computer skills
Tolerance and ambivalence
Effective communication abilities
Broad educational background
Advanced, socially sophisticated verbal skills
Fast-learning capabilities (of different domains)
Must understand organizations and individuals
Wide experience in knowledge engineering
Intelligence
Empathy and patience
Persistence (Tekun)
Logical thinking
Versatility and inventiveness
Self-confidence

Knowledge Acquisition Solutions

Knowledge Models

Decompose the world into

knowledge objects

Psychological
Techniques

Knowledge Objects - Example

"Comparing the
Kob
Kob to a similar
engin
e
engine,
SN the SN4,
4 the Kob
I'd say
a lot
waslighte
a lot smaller
r
and lighter.
I'd
characterise
hig the
h low
Kob
as having
high
efficien
cy
efficiency,
and
low
quie
emission
s
emissions
and t
quiet..
it's a quiet
engine".

Concepts:
Kob, engine, SN4
Relations:
a lot smaller, lighter
Values:
high efficiency, low
emissions, quiet

Knowledge Objects

Knowledge Models - Example

Kob
engin
e

SN4-100
SN4

concepts

SN4-200

reports
documents
specifications
lighter than
relations

smaller than
is a
composed of

design specs
test specs

Knowledge Models Further Examples

Semantic Network

Kob
smaller
than

SN4

lighte
r
than

Frame

Kob
efficiency

high

emissions

low

noise level

quiet

From Expert to Repository

Knowledge
Objects

Knowledge
Models

Knowledge
Repository

Proses Akuisisi Pengetahuan

(ES Development)

Domain Expert

Expert
System

Knowledge
Engineer

Question

Answers

Results

Knowledge

Basic Acquisition Process

Elicit knowledge
from experts

Documentation

Transcript

Perform knowledge
analysis

Knowledge objects

Construct knowledge
models
Knowledge models

Populate repository

Acquisition Techniques - Examples

Protocol Generation Techniques
Laddering
Protocol Analysis
Constrained Tasks
Process Mapping
Concept Sorting
Repertory Grid/Matrix-based

Protocol Generation Techniques

Interview

Unstructured Interview
Semi-Structured Interview
Structured Interview

Commentary
Teach Back
Observation

Laddering

Acquisition Techniques - Examples

Laddering
Protocol Analysis
Constrained Tasks
Process Mapping
Concept Sorting
Repertory Grid

Concept Ladder

Composition Ladder

Decision Ladder

Attribute Ladder

Procces Ladder

Acquisition TechniquesProtocol
- Examples
Analysis

Laddering
Protocol Analysis
Constrained Tasks
Process Mapping
Concept Sorting
Repertory

Expert provides a
running commentary
as they perform a task.
This is recorded and
transcribed.

Acquisition Techniques
Examples
Constrained Tasks
Laddering

Expert tries to guess

reasons for turning
one way or the other.

Protocol Analysis
Constrained Tasks
Process Mapping
Concept Sorting
Repertory Grid

?
?
The questions asked give
vital knowledge.

Acquisition Techniques
Examples
Process Mapping
look-up max-ht-ATC

look-up max-ht-Wx

Laddering

compute max CAP ht

look-up max-ht-airframe
compute CAP ht candidates
select lowest of max CAP ht candidates
propose CAP ht
compute min CAP ht

Protocol Analysis

compute optimal CAP ht

select lowest CAP ht candidates

select CAP ht
select highest CAP ht candidates

calc expected detection range

Constrained Tasks

calc max-range-time

calc max-range-comms
calc CAP max range
calc max-range-fuel

calc max-range-sector

Process Mapping

select lowest fromCAP max range

candidates
calculate CAP range candidates
propose CAP range

calc CAP min-range-ATC

calculate CAP position

propose horizontal CAP position

calc CAP min-range-safety

calc CAP min range
calc CAP intercept distance

Concept Sorting

calc CAP min-range-intercept

calc sea return distance

calc CAP optimal range
determine max sea return distance

Repertory Grid

select CAP range

calculate initial sectors

determine CAP sectors
modify initial sectors

decide 8 or 0
select orientation
decide CAP orientation angle

check OPTASK CAP pos based on fuel

check OPTASK CAP pos based on time

check validity of OPTASK CAP pos
check OPTASK CAP pos based on
routing

query OPTASK CAP ht/ pos

Acquisition Techniques
Examples
Concept Sorting
Laddering
Protocol Analysis
Constrained Tasks
Process Mapping
Concept Sorting
Repertory Grid

avicide
avicide
ss

mollusicide
mollusicide
ss

piscicide
piscicide
ss
repellen
repellen
tsts

slimicid
slimicid
es
es
disinfectan
disinfectan
tsts

Expert repeatedly sorts

cards, images or
objects into piles based
on similar properties.

Acquisition Techniques
Examples
Repertory Grid
Laddering
Protocol Analysis
Constrained Tasks
Process Mapping
Concept Sorting
Repertory Grid

Uncertainty in Knowledge Acquisition

Expert more than one
Incomplete data/Information
Difference Imprecise of Language (small, medium,

high, enough)
e.g :
Tiga orang pakar diminta menilai berapa usia seseorang masih dianggap
produktif
Tiga orang pakar diminta menilai apakah pendapatan seorang menteri
terlalu kecil , cukup , besar atau sangat besar

Pendekatan yang dilakukan

Fuzzy

Identifikasi jika terdapat variabel yang bersifat Fuzzy

Menggunakan Fuzzy Inference System

Certainty Factor

Melakukan akuisisi dengan mengidentifikasi impresi bahasa

yang disampaikan pakar (Misal: pasti, mungkin, cukup, kirakira dll)
Dari impresi bahasa tersebut dikonversi ke dalam CF

Bayes

Berbasis statistik dan konsep peluang bersyarat

Nave Bayes
Merupakan teknik pengambilan keputusan berbasis

statistik
Konsep peluang bersyarat
Nave Bayes : Sebagai classifier
Contoh kasus [klik]

Certainty Factor
MB MD
CF = CF (evidence) * CF (Rule)
Kombinasi (CF)

Acquisition Software - PCPACK

Why have so many techniques? :-\

Comparison of KA Techniques

1. Conduct an initial interview with the expert in order to (a) scope

what knowledge is to be acquired, (b) determine what purpose the
knowledge is to be put, (c) gain some understanding of key
terminology, and (d) build a rapport with the expert. This interview (as
with all session with experts) is recorded on either audiotape or
videotape.
2. Transcribe the initial interview and analyse the resulting protocol.
Create a concept ladder of the resulting knowledge to provide a
broad representation of the knowledge in the domain. Use the ladder
to produce a set of questions which cover the essential issues across
the domain and which serve the goals of the knowledge acquisition
project.
3. Conduct a semi-structured interview with the expert using the preprepared questions to provide structure and focus.
4. Transcribe the semi-structured interview and analyse the resulting
protocol for the knowledge types present. Typically these would be
concepts, attributes, values, relationships, tasks and rules.
5. Represent these knowledge elements using the most appropriate
knowledge models, e.g. ladders, grids, network diagrams,
hypertext, etc. In addition, document anecdotes, illustrations and
explanations in a structured manner using hypertext and template
headings.
6. Use the resulting knowledge models and structured text with
contrived techniques such as laddering, think aloud problem-solving,

Summary

Knowledge Acquisition = Identifying

and capturing knowledge assets to
populate a knowledge repository.

A major part of
Knowledge Acquisition
is capturing knowledge
from experts.

Capturing knowledge
from experts is difficult
and can be very time
consuming, costly and
inefficient.

Capturing knowledge from

experts is made costeffective and efficient by
using knowledge models
and special psychological
techniques.

Use of knowledge models

and specialised software
tools provides efficient
transfer to the knowledge
repository.