Implementing an Ontology-Based Knowledge Management System in the Korean Financial Firm Environment

walkingceilInternet and Web Development

Oct 22, 2013 (3 years and 11 months ago)

79 views

Implementing an Ontology
-
Based Knowledge Management System in the

Korean Financial Firm Environment



Hyun Hee Kim

Department of Library and Information Science, Myongji University,
Seoul, Korea.
Email:
kimhh@mju.ac.kr


Soo Young Rieh

School of Infor
mation, University of Michigan, Ann Arbor, MI 48109
-
1092. Email: rieh@umich.edu



Tae Kyoung Ahn

The Korea Institute for International Economic Policy, Seoul, Korea. Email: tkahn@kiep.go.kr


Woo Kwon Chang

IFK,
Myongji University,
Seoul, Korea.
Email: wk
1961@mju.ac.kr




This paper reports
one

part of a larger research
project
that

design
s

and implement
s

an
ontology
-
based knowledge management system
which
makes

knowledge assets intelligently
accessible
to
Korean financial firms. Th
e

research
develops

a
n ontology model co
nsisting

of an information ontology, a competency
ontology, a product ontology
, and a knowledge
map
. This paper introduces the ontology
model

by

illustrating

the
four components and reports
on the implementation and evaluation of the
in
formation ontology for
the
search
ing

of
web
resources. Based on the content analysis of
eight international bank web sites, a pilot system
of information
ontology

for

web resources


consisting of

Publication, Project, Member,
Person, and Organization

was
constructed
.

Th
e

pilot system
includes

databases and ontology
-
based search engines. To evaluate the pilot
system,
a

comparative experiment

was
conducted
in which the performance of
an
ontology
-
based system was compared with
that
of
web search engines. Th
e results indicate
that
the
ontology
-
based system can be used not
only to improve precision but also to reduce
search time.



Introduction


Knowledge Management (KM) can be defined as
any process or practice of creating, sharing, and
applying
knowledge

(e
xplicit or implicit)

wherever it resides for
better
performance

in organizations (Swan et al., 1999).
There is
a
consensus among organizations that
knowledge
is an essential asset for success and survival in an
increasingly competitive and global market.

However,
in most organizations
KM
applications

have not
to date
proven

largely successful
, arguably
because Knowledge
Management

(KM)

systems

have paid too much attention
to technical solutions
and too

little
to

the value of well
-
organized knowledge within

the system.

Most vital organizational knowledge
can be seen as

resid
ing, for instance,
in the skills and memory of the
organizational members. Thus successful knowledge
management
need
s

to

consider not only technical
but also
social aspects
.

The t
ime
is
apparently

right for
information science researchers to
investigate

how to
represent shared knowledge
as well as

how to

create an
organization culture that encourages knowledge sharing.

In this study,
ontologies are employed for both
knowledge
-
sharing
a
nd semantic

web searches.
The

research

goal of this research is threefold:



To design
the four components of

information
ontology, competency ontology, product ontology
,
and knowledge map



To implement a pilot system based on these
proposed
components



To
ev
aluate

the pilot system in terms of relevance,
customer satisfaction, and staff satisfaction.

This paper is structured first
to

describe
the

m
odel of
o
ntology
-
b
ased
k
nowledge
m
anagement
, second to
introduce
a pilot system that implements
the

information
ontology for
semantic
web
searches, and
third

to

illustrate


the evaluation study of
the

pilot system.




What is Ontology?


Ontology can be simply defined as
a formal, explicit
specification of a shared conceptualization

(
Gruber, 1993).


However,
ontol
ogy
has been

often
employed
without
an

attempt
at

standard definitions
;

worse, it has been used

to refer to

different things in different fields. In
library
and
information science, ontology
has
often mean
t

glossaries, data, thesauri, taxonomies, and f
ormal
ontologies and inference. Thesauri, taxonomies and
ontologies all have in common a method of
relat
ing

terms
in a controlled vocabulary via

a semantic

hierarchy

and
associative relationships

(Soergel, 1999).



There are, however, some differences
.
A
t
axonomy

is a semantic hierarchy in which information entities are
arranged by

hierarchical relationship
s

whereas a thesaurus
deals with the relationship
s

between

terms
that are
structured taxonomically (Daconta, Obrst & Smith, 2003).
That is
,

a
thesaurus

is
a combined form of taxonomy

and
semantic relations of terms.
A

formal ontology
means

the complex semantics of concepts and the relations
among concepts, their properties, attributes, values,
constraints, and
rules. More importantly,
ontologies as
expr
essed in an ontology representation language such as
Web Ontology Language (OWL) can
add values
to
taxonomies

or thesauri through deeper semantics
(Kwasnik, 1999; Qin & Paling 2001).

According to
Ding and Foo (2002), t
he concept of deeper semantics can
i
mply deeper levels of hierarchy, enriched relationships
between concepts, conjunction and disjunction of various
concepts, and formulation of
inference

rules.
Another
difference is that thesauri

and taxonomies
are intended

for
humans whereas ontologies can

be used by software
agents for knowledge
-
processing

as well as

by humans for
knowledge
-
sharing.



Related Literature


Ontology
-
B
ased
Information Representation and
Retrieval



Apparently
some researchers do not distinguish
ontology
-
based information ret
rieval from thesaurus
-
based
retrieval
.

T
he distinction between these two
seems
, however, to be of

importan
ce
.
WorldNet

(Fellbaum, 1998) is one example of a linguistic ontology;
however,
it can be considered

as
an online lexical
reference system rather th
an a formal ontology. With a
design inspired by current psycholinguistic theories of
human lexical memory, WorldNet contains
about
100,000
word meanings taxonomically organized
.



Unlike

taxonomies and thesauri, ontologies provide
both better semantic re
presentation and machine
-

understandable representation of knowledge. Thus
,
a
number of
ontologies were constructed based upon
existing thesauri
. One example is the Unified Medical
Language System (UMLS)
developed

by the National
Library Medicine (Kashya
p & Borgida, 2003). The
UMLS consists of Metathesaurus (MT), Semantic
Network

(SN), and a collection of lexical tools. As a
semi
-
formal ontology, the SN consisting of semantic
types and relationships
is
used to categorize MT concepts.
Another example re
sides in the study
of
Qin and Paling
(2001) that converted the controlled vocabulary of the
Gateway to Educational Materials (GEM) into an
ontology. The demand to convert
controlled

vocabularies
into ontologies is due to the limited expressive power of
con
trolled vocabulary and semantic
-
searching through the
Internet and intranet.


Of the studies done on ontology
-
based information
retrieval, Hyvonen, Styrman and Saarela (2002) built such
a system in which images were annotated according to
ontologies
. Th
ei
r

system offered
the same
conceptualization to facilitate focused image retrieval
using correct terminology.

Sure and Iosif (2002)
compared

two ontology
-
based search tools with a typical
keyword
-
based search tool in terms of search time,
mistake
-
making, u
sefulness, and tool development and
maintenance
. The first ontology
-
based search engine
was used to obtain a picture of the information available
in the system knowledge base

while

the second one was
used to give users semantic context. The search result
s

revealed that
ontology
-
based

tools were generally at least
as good as the keyword
-
based tools and, to an extent,
even superior.


Ontology
-
B
ased
Knowledge Management



Abou
-
Zeid (2003) proposed
an
ontology engineering
process model based on the premise t
hat ontology
development is a special
application

of Nonaka’s
knowledge creation model.

Nonaka

s

model views
the
ontology engineering process as
a
spiral in which
interaction between

ontology stakeholders


tacit
knowledge and explicit knowledge
is contin
uous and
dynamic
.


Vasconcelos (2001) applied knowledge management
and knowledge engineering techniques to the design of
information systems, especially organizational memory.
For this purpose he
used ontologies to
design, implement,
and evaluate a Grou
p Memory System.


Reimer et al. (2003) described two case studies
conducted by the Swiss Life Insurance Group
to
prov
e

the
practical applicability and superiority of ontology
-
based
knowledge management over classical approaches based
on text
-
retrieval te
chnologies. The first case study
,

in
the domain of skills management
,

used manually
-
constructed ontologies on the subjects of skills, job
functions, and education. The system’s purpose was to
support
the

identif
ication of

employees with particular
skills
. The second case study aimed to improve content
-
oriented access to passages of 1000
-
page documents
through the use of

an ontology consisting of 1500
concepts linked by 47,000 weighted semantic associations.




Minghong et al. (1999) proposed an ontology
-
ba
sed
competence system which facilitates the location of
appropriate contact persons for business tasks requiring
specific knowledge, experiences, or skills.

Sure (2003)
presented
a number of

semantic web methods and
technologies and showed their applicabi
lity to practical
knowledge management in corporate intranets and in the
web.
Quan, Huynh, and Karger (2003) investigated the
knowledge management system (Haystack) built on RDF
to create, organize, and visualize personal knowledge
such as e
-
mails, documen
ts, tasks, contacts, meetings, and
other information.


Even though previous studies on ontology
-
based
knowledge management ha
ve

proven to be active and
productive, few studies have

as

yet discussed an
integrated ontology
-
based knowledge management
syste
m focusing not only on search
es

of web resources
and
company

documents but also on support for
identifying employees with particular skills and
appropriate company products.



Conceptual Framework



The goal of this study is to design and implement
an

o
ntology
-
based knowledge management system
adapted
to the environment of
Korean financial firm
s, m
ost

of
whose

primary objective is

to provide customer
s

with
a

full range of

the most modern banking products and
services by

implementing the latest advances
i
n

information technology, by

developing and enhancing
business processes, and by

continuously improving
service quality. To achieve this objective, staff skill
management (competency), product development, and
information management appear to be the mos
t important
concepts.


As shown in Figure 1, the proposed ontology
-
based
model includes three
ontologies and an integrated search
engine
:

an information ontology, a competency ontology,
a product ontology
, and a knowledge map
. The
information ontology i
s used primarily
to

search both

web resources
and

financial firm documents. The
competency ontology is designed to identify employees
with particular skills as well as to
determine

the
kinds of
skills required to conduct tasks. The product ontology is
em
ployed
primarily
to

select products.
Finally
, the

knowledge map is the search engine employed to
integrate multiple resources such as structural,
unstructured, and expert databases.


Ontology Design

Method



The following section describes the methods
which
serve

as

the bases for designing three ontologies.




Figure 1. A Model of Ontology
-
Based Knowledge
Management



Information Ontology



For the information ontology for web resources, an
international banking firm was selected as a sample
domain. T
o design this ontology, content analysis was
conducted on

the web sites of

eight international bank
s
:
World Bank, ADB, OECD, EBRD, IFC, IMF, WTO, and
APEC.
On

content analysis
completion
,
one
interview
w
as

conducted with a domain expert in an economic
rese
arch institute.


Competency Ontology



In

design
ing

the competency ontology, a preliminary
content analysis
was

conducted
using

ten Korean bank
web sites and documents.
S
urveys and interviews with
the employees of

the

financial firms are in progress
:

a

total of about 60 questionnaires will be mailed to bank
employees

and

i
nterviews are in progress to identify the
kinds of skills that employees need for their work tasks.


Product Ontology



For product ontology, fifteen Korean bank web sites
and thei
r product databases will be analyzed and
compared. Once the analyses of websites are completed,
interviews will be conducted both with the

financial firm

employees and clients.



Ontology
-
Driven Modeling


Ontology

Design Process



The ontology

design pro
cess follows the widely
-
accepted five
-
stage methodological approach (
Uschold &


Gruninger, 1996; Vasconcelos, 2001)
:

(1) identification of
purpose and scope, (2) knowledge acquisition and
conceptualization, (3) integration and reuse of other
ontologies, (4)

formal specification, and (5) evaluation
and documentation. The design steps for building
ontologies are described below.


1) Ontology Purpose and Scope


This stage aims to define the purpose and scope of the
ontology, describing its use, its users, and
the scope of the
ontology. The proposed ontologies are constructed to
allow employers, managers, clients, librarians, and
general users to
search
semantically
for

information.


2) Knowledge Acquisition and Conceptualization


The process of acquiring know
ledge from a given
domain is described in this stage. In th
e

study, in order
to collect glossaries of concepts (classes) for the domains

content analyses are
first

performed on web sites and
company databases.
Next
, expert interviews and surveys
are con
ducted.
Finally
, domain concepts, instances,
relations, and properties are identified, presented, and
associated with domain terms.


3) Ontology Integration


Other ontologies
can be

employed

in the process of
building a new ontology. To obtain
a degree
of

uniformity across ontologies, definitions from other
ontologies
will be

reused. Several ontologies including
the enterprise
ontology

(
http://www.aiai.ed.ac.uk/

project/

enterprise/
enterpr
ise/ontology.html)
will be

consulted
in
the

design
of the
three ontologies.



4) Concept Description and Formal Specification


This stage proposes that ontologies should be formally
represented using an ontology language. OWL is used to
codify
three
on
tologies

given that

it is currently the best
-
known ontology language for the semantic web.

This
stage involves the formalization of each term and the
constraints used by the ontology. Terms are represented
through classes, relations, functions, and instan
ces.


5) Evaluation and Documentation


Ontology validation and verification are accomplished
through the application of a set of guidelines which look
for incompleteness, inconsistencies, and redundancies
(Gomez
-
Perez, 1995).
However,

direct evaluation
of
ontologies is difficult

and so
to evaluate
them

three
methods including comparative studies and usabilities
will be

employed.



Ontologies



1)
Information Ontology

The information ontology can be
divided

into two
categories: the ontology for web
resource
s

and the
ontology for company documents.
The o
ntology for web
resource

search
contains five subontologies including
Publication, Project, Person, Member, and Organization
while
t
he ontology for company document
search

contains
only

Document subont
ology as presented in
Figure 2.


Figure 2. Overview of Information Ontology



The relationship between two
sub
ontologies is
presented by solid lines. For example,
the
Publication
sub
ontology is related to
the
Project
sub
ontology because
some
projects
produce publications as the results of
research.
The D
ocument
sub
ontology c
an

be connected
to
the

Publication
sub
ontology when
internal
documents
are uploaded to web servers for the public.


The Is
-
A relation of Figure 2 indicates that one
concept is a

subclass of another, mean
ing

that the
collection of Is
-
A arcs specifies a categorization hierarchy.
Therefore, an
I
ndividual
M
ember is a subclass of
M
ember
.




T
able 1.
The Overview of
Publication
Subo
ntology















Publication (org_name, pubCode, pubAuthor, pubTitle,
pubLan, pubYear, pubResearch_topic,
pubResearch_counrty_region, has_URL)




Serial (pubFrequency)


J
ournalArticle (journalTitle)


Statistics (code)


WorkingPaper (seriesStat)


DiscussionPaper (seriesStat)


Outlook ()

Article ()

Conference (conferenceTitle)

………………………

UnofficialPublication ()





Table 1 presents
the

overview
of Publication
sub
ontology. This
sub
ontology include
s

metadata
elements to enhance the representation and retrieval of
publication resources.
In this subontology, the
Publication class is

a
top
-
level class

composed of nine

properties all of which are inh
erent to all of the subclasses
of the Publication class such as Serial and
UnofficialPublication.
The c
onferenceTitle property is
applied only to
the
Conference class.


2) Competency Ontology


The competency ontology is designed based on Stader
and Macin
tosh’
s

(1999) competency taxonomy. Th
is

taxonomy includes primitive competencies and
application areas. To design the competency ontology

adapted to the environment of

K
orean financial firm
s
,


some primitive competencies are deleted and new
primitive

competencies

added. Modifications are also
made in
the
application areas.


Figure 3. Overview of Competency Ontology


The main classes of Figure 3 are Competency and
Entity
. T
he Competency allows the representation of
different levels of competency gran
ularity through the
creation of subclasses of competencies
, and

the Entity
describes the different application areas in which a
specific competency can be applied. The two hierarchies
are combined
by

a set of competency relations such as
has
-
skill
-
of and h
as
-
experience
-
of

to allow the
combination of terms between hierarchies. For example,
to describe
the

specific domain expert

who

has
-
experience
-
of (competency relation) Dealing (application
area)
in

Foreign Exchange (primitive competency) as
formal notatio
n, has
-
experience
-
of (Foreign Exchange,
Dealing) is employed.


3) Product Ontology


As shown in
Table 2
, the product ontology has
the
four
main classes
of

Deposits, Trust, Investment Trust, and
Loans. The Deposit class has nineteen subclasses
, and

a
mong

them Installment Savings Deposit
s

and
Time
Deposits

have their own subclasses.

This product
ontology targets general clients
, but c
ustomized products
are
ex
cluded in this ontology because
of the

difficult
y of

fix
ing

their classes as well as their prope
rties.


Table 2. Overview of Product Ontology























Ontologies in OWL


The three ontologies

of information, competency, and
product

are represented in three OWL files. Here,
classes and properties are main components of OWL
language. A
class defines a group of individuals that
belong together because they share
certain

properties.
Classes can be organized in a specialization hierarchy
using SubClassOf. The first example
in

Table 3
indicates that
a
Researcher is a subclass of Person.


Properties are determined based on whether they
relate individuals to individuals (ObjectProperties) or
individuals to datatypes (DatatypeProperties). In Table 3,
Org_type

as a

property of
the
Organization class is
defined as an object property where the

Org_type
property ties an Organization to an OrganizationType.

Deposits















Demand Deposit
s











.........................




MMDA: Money Market…
.

Corporate Savings Deposit
s

Installment

S
avings Deposit
s



household installment deposit
s


…………

General Installment Deposit
s







Scholarship Deposit
s













Long Term Savings


.

Time Deposits


general time deposit
s




……………
..

Trust
s

O
pen Type Money Trust
s
.

Specified Money Trust
s

Personal Pension Trust
s

Investment Trust
s

Beneficiary Certificates

Off Shore Mutual Fund

Loans

Auto Loan
s

Loan Secured by Deposit
s


.
..


………………



Table 3. Classes and Properties

<!
--
subClassOf

>

<owl:Class rdf:ID="Researcher">


<rdfs:subClassOf rdf:resource=="#Person"/>

</owl:Class>



<!
--

ObjectProperty
--
>

<owl
:ObjectProperty rdf:ID="Org_type">


<rdfs:domain rdf:resource="#Organization"/>


<rdfs:range rdf:resource="#OrgnizationType"/>

</owl:ObjectProperty>



Knowledge Map



The knowledge map is
a

search engine that integrates
multiple resources su
ch as structural, unstructured, and
expert databases. Based on the above
-
mentioned product
and information ontologies, the structural databases
include product, web resources
,

and internal documents.
Conversely, the unstructured databases have codified
i
mplicit knowledge to be stored in web pages or
Lotus
Notes

(Davenport & Prusak, 1998).


The

expert databases
are

added in order to

access
experts with tacit knowledge and are designed based upon
the competency ontology. Figure 4 gives the overview
of t
he knowledge map.


Figure
4
. Overview of
Knowledge Map



S
ystem Architecture




Figure
5

provides
an

overview of
the
system
architecture. The ontology
-
based module has two
components: databases and ontology
-
based search
engines.


Figure
5
. System Arc
hitecture



The databases have two kinds of files
: the

ontology
file
s

in OWL and
the
RDBMS files.
OWL ontology files
are employed to design tables and their attributes in
the
RDBMS files. To better understand data structures, t
he
Resource Description Frame
work (
RDF) model is first

constructed
.
Based on the RDF model,
annotated web
resources are stored in RDF triple table
s
.

The Windows
NT server is utilized as the system’s server
.



Database



The proposed system has two kinds of database:
OWL
ontology fil
e
s

and RDBMS files.


1)
OWL
Ontology File
s



Each ontology has
an
ontology file represented in the
OWL language, resulting
in
three OWL ontology files.
For example, the information ontology has six top
-
level
classes including Publication and Document.

The full
version of the information ontology has 3
5

classes and
64

properties.


2) RDBMS Files


According to
OWL ontology files
, database,
web
resources, and internal documents are annotated manually
and then, based on the RDF model, their annotated dat
a
are stored in the RDF triple tables of the RDBMS.

Figure 6 shows the RDF graph which indicates that there
is an article identified by the web site address shown in
Figure 6 as well as that this resource has five properties
with accompanying values.




F
igure
6
. RDF Model




Based on this RDF model, annotated resources are
stored in the RDF triple table shown in Table 4.


Table
4
. RDF Triple Table

Resource

Property

Value

http://www.imf.org/external/ftp

/
wp/2002/wp02204.pdf

http://purl.org/dc/

elements/1.1/

creator

Leigh, Danie

http://www.imf.org/external/ftp

/
wp/2002/wp02204.pdf

http://purl.org/dc/

elements/1.1/

title

Exchange
Rate Pass
-
Through in
Turkey

http://www.imf.org/external/ftp

/
wp/2002/wp02204.pdf

http://purl.org/dc/

elements
/1.1/

type (1)

Serial


http://www.imf.org/external/ftp

/
wp/2002/wp02204.pdf

http://purl.org/dc/

elements/1.1/

type (2)

WorkingPaper

http://www.imf.org/external/ftp

/
wp/2002/wp02204.pdf

http://www.daml.ri.cmu
.

edu/ont
/
homework
/atlas

-
cmu
.daml/
organization

IMF


http://www
.imf.org/external/ftp

/
wp/2002/wp02204.pdf

http://purl.org/dc/

elements/1.1/

date

2002




Ontology
-
Based Search Engin
e



Based on the RDBMS files, computer programs
are
created to

allow
the

conduct
ing of

onto
logy
-
based
searches. MS SQL RDBMS has been
here
employed.
However, the RDBMS
reveals

some limitations in
representing hierarchy relation between classes or
properties and in defining inference rules. Search
programs and RDBMS records have therefore been
designed to solve
such

problems. For example
,

to allow
users to do hierarchy searching in a Publication DBMS
file, two fields for a document type were created. For
first
-
level classes such as Serial only the first fields used
for
the

first
-
level classes ar
e to be searched. For second
-
level classes such as JournalArticle only the second fields
used for
the

second
-
level classes are to be searched.



System Implementation of Information
Ontology for Web Resources



A full range of pilot systems is under deve
lopment.
This section describes an example of
an

Information
ontology search. Figure
7

shows
the

first screen of the
ontology
-
based knowledge management system.





Figure
7
. Search Screen



After
a selection of

information ontology

is made

from

the le
ft frame of Figure
8
, the right frame shows a
search mode for the information ontology. In the search
mode, a class (category)

is selected
from a drop
-
down list
and the select button

then pressed
, at which the system
responds
by producing

a set of proper
ties applicable to
that class. Applicable properties are inheritable; thus
any properties that apply to an ancestor of the selected
class are also included in the set. The following shows
how to select a class
as well as how to

search databases
using an
example.


After
selection of

the WorkingPaper class as a
category and “all of international organizations” as target
organizations, the system responds
by producing

the


seven properties of the WorkingPaper class as shown in
Figure
8
. The property list a
llows the issuing of a query.
For this search, “economic effects of ageing” is selected
as
a

subject code, “English” as
a

language code, and “all”
as
a

country/area code
, after which the system displays
the search result
s
. The subject code was construct
ed
based on the topic taxonomy of OECD home page
(
www.oecd.org
).

Figure
9

shows the full record of
Marcos’s paper
upon

the clicking of one short record.
By selecting a hyperlinked “Resource_URL”
field
, full
-
text docume
nts in PDF format

can be accessed
.




F
igure
8
. Properties of the WorkingPaper Class




Figure
9
. Result Screen (Publication)






System Evaluation of Information Ontology
for Web Resources



A

comparative experiment was conducted

t
o evaluate
the info
rmation ontology for web resources
.

The
performance of
the
ontology
-
based system was compared
with
that of
web search engines in terms of relevance and
search time. Ten researchers from an economic research
institution were recruited in October, 2002
, to c
onduct
experiments
in the
researchers
’ office
s

except
on
two
occasions
when both of which

were
conducted

in the
library. Before the participants conducted
their
searches,
they were
provided

by the experimenter with a
half
-
hour
presentation about the system
.


The
participants

were given a list of twenty tasks and
asked to perform both on a search engine
of their

cho
ice

and
the

ontology
-
based pilot system.

The tasks included
answering

questions about
the locating of

scholarly
literature, statistical data,
conference information, news,
people search
es,

and project search
es
.


The effectiveness of the pilot system was measured
with respect to relevance and

completion

time
for

each
task. Relevance was measured with respect to
the
precision of the top 15 results

for each task. For each 15
results, the
researcher selected a

score of relevance on
a

0
-
5

scale

in which 0=don’t know, 1=very irrelevant,
2=irrelevant, 3=neutral, 4=relevant,

and

5=very relevant.
The average relevance scores from 10 participants for
ea
ch task were
then
calculated and compared across 20
tasks as shown in Figure 10.


The difference
in

relevance scores between

the

two
systems appeared to be relatively higher when the
participants

were looking for information about people
(Q1, Q10, Q16),
conference information (Q6), image (Q8,
Q9, Q19),
and
news information (Q18) while the
difference was lower in the cases of scholarly literature
(Q5, Q7, Q12, Q13, Q17).


The results also show that
,

o
verall
,

the average
relevance score of the ontology
-
ba
sed system (4.53) was
higher than that of web search engines (2.51). The
average search time
for

ontology
-
based search
es

and
for

w
eb search
es

was compared: overall 1.96 minutes were
taken when using the ontology
-
based system and 4.74
minutes were taken wh
en using general
w
eb search
engines.


As shown in Figure 11, the difference in search time
was greater when the
participants

were looking for
information about images (Q8), project
s

(Q11), people
(Q16), and organization information (Q20) while the
differe
nce was smaller for the task
s

on

scholarly literature
(Q7, Q13).





Figure 10. Relevance
Comparison of
Ontology Search and Internet Search


Figure 11. Search Time Comparison of Ontology Searc
h and Internet Search




Conclusion



There are several issues in the information retrieval
area.
One of them is that web search engines produce an
excess of search results and another is that
many
retrieved
sites are irrelevant. These problems are at
tributable to the
fact that common information
-
retrieval techniques rely
either on specific encoding of available information or
simple full
-
text analysis, both of which lead to limitations
such as
ambiguity

of
word

meanings and vocabulary
inconsistency

be
tween texts and users. T
o address

the
limitations, this study designed and implemented the
ontology
-
based web retrieval system in which ontologies
were utilized to add semantics to web pages for use in
semantic web searches (
Berners
-
Lee
&

Fischetti
, 1999)
.


T
o evaluate the
pilot system
,
t
he performance of
an
ontology
-
based system was compared
to

that of
web
search engines
.
The results
revealed

that the average
relevance score of the ontology
-
based system
was
higher
than that of web search engines. The ave
rage search time
taken

for

ontology
-
based search
es was much shorter than
that
taken

for w
eb search
es.

The
greatest

d
ifficulty

in th
is
ontology
-
based
approach is the extra work
necessitated by the

annotati
on
of web sites based on ontologies
.
D
ue to the exp
ense of
annotating resources, it appears that
certain

domains such
as
the
business sector might be more appropriate than
others for ontology applications.

The future of this project
will entail implementing

the following ontologies using the same methodolo
gy
as

information ontology for web resources: the information
ontology for financial firm documents, the competency
ontology, and the product ontology. That is,
the three

will
be transformed into OWL ontology files and RDBMS
files. The RDBMS files will b
e constructed based on
OWL files and RDF models. Data will be extracted
from databases and web sites of financial firms and then
input into records of tables. The final pilot system will
feature

these three ontology subsystems as well as the

knowledge m
ap
for

integrating multiple resources. Th
e

system will be evaluated in terms of relevance, financial


firm employee satisfaction, and financial firm client
satisfaction.



T
he proposed information ontology was constructed
using expert implicit knowledge
as well as banking firm


databases.

Thus,
to make the proposed pilot system
update systematically, the environment
such as
communities of practice

will be

need
ed

to collect

implicit knowledge from the
financial firm
employees and
clients.

Additionally
,

deep annotation of

annotating
databases (Handschuh, Staab, & Volz, 2003) can be
utilized

for the web databases of the pilot system.



ACKNOWLEDGEMENTS


This work was supported by the Korea Research
Foundation Grant (KRF
-
2002
-
005
-
B20006)



REFERENCES


Abo
u
-
Zeid, E
.
S
.
(2003). What can ontologists learn from
knowledge management?

J
ournal

of Computer Information
Systems,
43
(
3
)
, 109
-
117
.


Berners
-
Lee
, T. &

Fischetti
, M. (1999).
Weaving the Web: t
he
o
riginal
d
esign and
u
ltimate
d
estiny of the World Wide Web
.

San Francisco: HarperSanFrancisco
.

Daconta, M., Obrst, L., & Smith, K. (2003).
The Semantic Web.

Indianapolis,
IN: Wiley

Publishing.

Davenport, T. H. & Prusak, L. (1998).
Working knowledge:
how organizations manage what they know.

Boston: Harvard
Busin
ess School Press.

Ding, Y.

&

Foo, S.
(
2002
)
. Ontology research and development
Part 1: A review of ontology generation.

Journal of
Information Science, 28
(2)
,

123
-
136.

Fellbaum, C. (E
d.
)

(
199
8)
.
WordNet: an electronic lexical
database.
Cambridge, MA: MIT
Press.

Gomez
-
Perez, A. (1995).
Some ideas and examples to evaluate
ontologies
,

T
he 11
th

Conference on Artificial Intelligence for
Applications

(pp.299
-
305).


Gruber, T. R. (1993). A translation approach to portable
ontology specifications.
Knowledge

Ac
quisition,
5, 199
-
220.

Handschuh,

S.,
Staab,

S., &

Volz
, R
.
(2003).
O
n Deep
Annotation
.


Proceedings of the 12th International World
Wide Web Conference
, Budapest, Hungary, May 20
-
24, 2003

(pp. 431
-
438)
.

Hyv
on
en, E
.
, Styrman
, A.
& Saarela
, S.
(
2002
)
.

Ont
ology
-
based image retrieval.
In

Ero Hyvonen & Mika Klemettinen
(Eds),
Towards the semantic Web and Web services
Proceedings
of
the XML Finland 2002 conference

(pp. 15
-
27).
Hiit Publications
.

Kashyap, V. & Borgida, A. (2003). Representing the UMLS
semantic

network using OWL.
In Fensel, D. et al (Eds),

2nd International Semantic Web Conference
(pp.1
-
16).

Kwasnik, B. (1999). The role of classification in
knowledge

representation and discovery.
Library Trends
,

48(1)
, 22
-
47
.

Minghong, L
.

et al. (1999). A
competence knowledge base
system as part
of the

organizational
memory XPS
-
99:
knowledge
-
based systems
, survey and future directions.

Lecture Notes in Computer Science, 1570
, 125
-
137.

Qin, J.
& Paling
, S. (2001). Converting a controlled vocabulary
into a
n ontology: the case of GEM.
Information Research 6(2)
.

Retrieved April 24, 2004, from:
http://informationr.net/ir/6
-
2/paper94.html.

Quan,
D
.
, Huynh
, D.
, & Karger
,

D
.

R.
(2003).
Haystack: A
Platform for Authoring End User Semantic Web Applications.

Proceed
ing of
The International Semantic Web Conference

(
pp.
738


753
).

Reimer et al. (2003). Ontology
-
based knowledge management
at work: the Swiss Life case studies, In Davies, J. et al (Eds),
Towards the semantic
web: Ontology
-
driven

knowledge
management

(pp
.197
-
218).
West Sussex, England:
John Wiley
& Sons.

Soergel, D. (1999). The rise of ontologies or the reinvention of
classification.
Journal of the American Society for Information
Science, 50(12),

1119
-
1120.

Stader, J
.

&

Macintosh, A
.

(1999).
Capability

modeling and
knowledge management. In
Applications and Innovations in
Intelligent Systems VII,
London:

Springer
-
Verlag

(
pp 33
-
50
)
.

Sure
, Y.
& Iosif
, V. (2002).

First
r
esults of a
s
emantic
w
eb
t
echnologies
e
valuation.


In Meersman et al
(Eds.),
Proceeding
s of the Common Industry Program
held in
co
njunction with Confederated I
nternational
C
onferences:

On the Move to Meaningful Internet Systems

(
CoopIS
, DOA,
and ODBASE 2002) (pp. 69
-
78).

Sure
,

Y
. (2003).
Methodology, tools & case studies for ontology
based
knowledge management.

Unpublished doctoral
dissertation
,
Karlsruhe University, Germany.

Swan J., Scarbrough, H., & Preston,

J. (1999).

Knowledge
management: A literature review
,
in issues in people

management
. Institute of Personnel and Development:
Leice
ster Warwick.

Uschold, M. & Gruninger, M.
(
1996
).

Ontologies: Principles,
methods and applications
.

The Knowledge Engineering Review,

11
(2)
,

93
-
136.

Vasconcelos, Jose Angelo Braga de.
(
2001
)
. An ontology
-
driven
organisational memory for managing group co
mpetencies.

Unpublished doctoral dissertation
, University of York.