K n o w l e d g e M a n a g e m e n t t h r o u g h O n t o l o g i e s V R i c h a r d B e n j a m i n s S W I U n i v e r s i t y o f A m s t e r d a m T h e N e t h e r l a n d s r i c h a r d s w i p s y u v a n l D i e t e r F e n s e l A I F B U n i v e r s i t y o f K a r l s r u h e G e r m a n y d f e a i f b u n i k a r l s r u h e d e A s u n c i o n G o m e z P e r e z D I A T e c h n i c a l U n i v e r s i t y o f M a d r i d S p a i n a s u n d e l i c i a s d i a u p m e s A b s t r a c t M o s t e n t e r p r i s e s a g r e e t h a t k n o w l e d g e i s a n e s s e n t i a l a s s e t f o r s u c c e s s a n d s u r v i v a l o n a i n c r e a s i n g l y c o m p e t i t i v e a n d g l o b a l m a r k e t T h i s a w a r e n e s s i s o n e o f t h e m a i n r e a s o n s f o r t h e e x p o n e n t i a l g r o w t h o f k n o w l e d g e m a n a g e m e n t i n t h e p a s t d e c a d e O u r a p p r o a c h t o k n o w l e d g e m a n a g e m e n t i s b a s e d o n o n t o l o g i e s a n d m a k e s k n o w l e d g e a s s e t s i n t e l l i g e n t l y a c c e s s i b l e t o p e o p l e i n o r g a n i z a t i o n s M o s t c o m p a n y v i t a l k n o w l e d g e r e s i d e s i n t h e h e a d s o f p e o p l e a n d t h u s s u c c e s s f u l k n o w l e d g e m a n a g e m e n t d o e s n o t o n l y c o n s i d e r t e c h n i c a l a s p e c t s b u t a l s o s o c i a l o n e s I n t h i s p a p e r w e d e s c r i b e a n a p p r o a c h t o i n t e l l i g e n t k n o w l e d g e m a n a g e m e n t t h a t e x p l i c i t l y t a k e s i n t o a c c o u n t t h e s o c i a l i s s u e s i n v o l v e d T h e p r o o f o f c o n c e p t i s g i v e n b y a l a r g e s c a l e i n i t i a t i v e i n v o l v i n g k n o w l e d g e m a n a g e m e n t o f a v i r t u a l o r g a n i z a t i o n I n t r o d u c t i o n A c c o r d i n g t o I n f o r m a t i o n W e e k A P H t h e b u s i n e s s p r o b l e m t h a t k n o w l e d g e m a n a g e m e n t i s d e s i g n e d t o s o l v e i s t h a t k n o w l e d g e a c q u i r e d t h r o u g h e x p e r i e n c e d o e s n t g e t r e u s e d b e c a u s e i t i s n t s h a r e d i n a f o r m a l w a y B e c a u s e t h i s c a n b e a n y k i n d o f k n o w l e d g e t a c i t d o c u m e n t e d p r o c e d u r a l e t c t h e t e r m k n o w l e d g e m a n a g e m e n t m a y r e f e r t o s u c h v a r i o u s t h i n g s T h e c o p y r i g h t o f t h i s p a p e r b e l o n g s t o t h e p a p e r s a u t h o r s P e r m i s s i o n t o c o p y w i t h o u t f e e a l l o r p a r t o f t h i s m a t e r i a l i s g r a n t e d p r o v i d e d t h a t t h e c o p i e s a r e n o t m a d e o r d i s t r i b u t e d f o r d i r e c t c o m m e r c i a l a d v a n t a g e P r o c o f t h e n d I n t C o n f o n P r a c t i c a l A s p e c t s o f K n o w l e d g e M a n a g e m e n t P A K M B a s e l S w i t z e r l a n d O c t U R e i m e r e d h t t p s u n s i t e i n f o r m a t i k r w t h a a c h e n d e P u b l i c a t i o n s C E U R W S V o l

lovemundaneManagement

Nov 6, 2013 (3 years and 9 months ago)

185 views


V Ric hard Benjamins Dieter F ensel Asunci on G omez P erez
SWI AIFB DIA
Univ ersit y of Amsterdam Univ ersit y of Karlsruhe T ec hnical Univ ersit y of Madrid
The Netherlands German y Spain
ric hardwisy v al dfeifbni arlsruhee asuneliciasiapms
ii O as corp orate memories and instincts
exp ert systems do cumen t managing systems learning
Abstract
organizations HvdSK etc
Kno wledge managemen t is not a pro duct in itself
Most en terprises agree that kno wledge is an
nor a solution that organizations can buy othehelf
essen tial asset for success and surviv al on a
or assem ble from v arious comp onen ts It is a pro cess
increasingly comp etitiv e and global mark et
implemen ted o v er a p erio d of time whic h has as m uc h
This a w areness is one of the main reasons for
to do with h uman relationships as it do es with business
the exp onen tial gro wth of kno wledge manage
practice and information tec hnology T The pro cess
men t in the past decade Our approac h to
of managing kno wledge in v olv es the follo wing actions
kno wledge managemen t is based on on tolo
Kno wledge gathering acquisition and collection
gies and mak es kno wledge assets in telligen tly
of the kno wledge to b e managed
accessible to p eople in organizations Most
Kno wledge organization and structuring imp os
compan yital kno wledge resides in the heads
ing a structure on the kno wledge acquired in order
of p eople and th us successful kno wledge man
to manage it ectiv ely
agemen t do es not only consider tec hnical as
Kno wledge reemen t correcting up dating
p ects but also so cial ones In this pap er w e
adding deleting kno wledge in short main tain
describ e an approac h to in telligen t kno wledge
ing kno wledge
managemen t that explicitly tak es in to accoun t
Kno wledge distribution bringing the kno wledge
the so cial issues in v olv ed The pro of of con
to the professionals who need it
cept is giv en b y a largecale initiativ e in v olv
ing kno wledge managemen t of a virtual orga
W e can distinguish b et w een t w o t yp es of kno wledge
nization
managemen t systems v ertical and horizon tal systems
V ertical systems are dev elop ed for one particular kind
In tro duction
of business situation Suc h systems are highly ectiv e
and ha v e pro v en their v alue Often v ertical systems
According to Information W eek PH he busi
are dev elop ed inside a compan y and are highly situa
ness problem that kno wledge managemen t is designed
tion sp eci Therefore suc h systems are of little v alue
to solv e is that kno wledge acquired through exp erience
for other business situations Horizon tal kno wledge
do esn get reused b ecause it isn shared in a formal
managemen t systems are general systems that can b e
w a y Because this can b e an y kind of kno wledge
applied to a v ariet y of business situations They are
tacit do cumen ted pro cedural etc the term kno wl
framew orks that can b e instan tiated to particular sit
edge managemen t ma y refer to suc h v arious things
uations ee PH for a discussion of e of suc h
The c opyright of this p ap er b elongs to the p ap er authors Per
systems Wincite In trasp ect ChannelManager Bac k
mission to c opy without fe e al l or p art of this material is gr ante d
W eb and Kno wledgeX
pr ovide d that the c opies ar e not made or distribute d for dir e ct
In this pap er w e presen t a horizon tal approac h to
c ommer cial advantage
kno wledge managemen t that is grounded in researc h
Pro c of the d In t Conf on Practical Asp ects of
on kno wledge engineering Kno wledge engineering is
Kno wledge Managemen t AKM
Basel Switzerland Oct Reimer ed
a ld that during the past y ears has b een con
h ttpunsitenformatikwthac heneublicationsEURS ol cerned with capturing analyzing organizing struc
V Benjamins D F ensel A G omez P er ez Annotating
Ontology building
turing represen ting manipulating and main taining
web pages
kno wledge in order to obtain in telligen t solutions for
Distributive or
Experts
Joint effort
hard problems BF O It is therefore no sur
centralized support
Users Users
ITers
prise that kno wledge engineering metho dologies and
<html>
tec hniques can b e of high v alue for kno wledge man
<a onto=
"page:
Ontology of
agemen t whic h is exactly concerned with the issues Annotated
employee">
subject matter
Web pages
</a>
men tioned ab o v e in a business en vironmen t AA
</html>
In order for our approac h to w ork in a partic
ular organization w e assume that it has an In
Intelligent
query answer
tranetxtranet or access to the In ternet and that
webcrawler
eac h mem b er of the organization has a bro wser In
addition the approac h requires that the kno wledge of
Figure The approac h
in terest is a v ailable in HTML pages on the net Man y
a senior manager ev en though this is no where stated
companies already ha v e an in tranet whic h is an easy
explicitly
to use infrastructure that giv es companies access to
Figure giv es a general o v erview of the approac h
a large v ariet y of In ternet tec hniques Therefore for
An on tology of the sub ject matter has to built whic h is
users already familiar with bro wsers our approac h has
used to c haracterize the sub ject matter to l the
a short learning curv e
on tology with instances An in telligen t w eb cra wler
This pap er is organized as follo ws In Section w e
receiv es a query in terms of the on tology consults the
outline the tec hnology underlying our approac h In
sub ject matter he instances in terprets them using
Section w e presen t an application of our approac h
the on tology and generates an answ er The instances
for a virtual organization the kno wledge acquisition
he actual kno wledge to b e managed are distributed
researc h comm unit y W e indicate ho w this case study
o v er diren t HTML pages f an in tranet or the In
relates to a business con text In Section w e iden
ternet
tify a n um b er of p ossible dangers to successful imple
men tation of kno wledge managemen t systems Finally
On tological Engineering
Section concludes the pap er b y putting it in con text
and relating it to the no wledge Chain of kno wledge
An on tology is a shared and common understanding of
managemen t
some domain that can b e comm unicated across p eople
and computers ru Gua UG vSW On
tologies can therefore b e shared and reused among dif
An on tologyased approac h to
feren t applications FR whic h is one of the main
kno wledge managemen t
reasons wh y on tologies are p opular no w ada ys An on
tology can b e deed as a formal explicit sp eciation
Our approac h comprises three main subtasks on
of a shared conceptualization ru Bor on
tological engineering to build an on tology of the sub
ceptualization refers to an abstract mo del of some
ject matter c haracterizing the kno wledge in terms
of the on tology and pro viding in telligen t access to phenomenon in the w orld b y ha ving iden tid the rel
the kno wledge In a sense this is reminiscen t of rela ev an t concepts of that phenomenon xplicit means
tional database tec hnology where the on tology w ould that the t yp e of concepts used and the constrain ts
corresp ond to the data mo del the c haracterization on their use are explicitly deed ormal refers to
w ould corresp ond to the instances ata con tained the fact that the on tology should b e mac hine read
in the database and access w ould tak e place through able hared rects the notion that an on tology
SQL W e will sho w ho w ev er that our approac h is sig captures consensual kno wledge that is it is not pri
nian tly diren t from cen tralized databases esp e v ate to some individual but accepted b y a group An
cially with resp ect to distributiv eness and in telligence on tology describ es the sub ject matter using the no
Our approac h captures distributiv e rather than cen tions of concepts instances relations functions and
tralized kno wledge The kno wledge is directly accessed axioms Concepts in the on tology are organized in tax
at its original lo cation n HTML pages rather than b e onomies through whic h inheritance mec hanisms can b e
separately input to a database The approac h allo ws to applied
isco v er kno wledge that is not explicitly kno wn but In order to come up with a consensual on tology of
that can b e deduced based on general kno wledge ap some domain it is imp ortan t that the p eople who ha v e
tured in the on tology F or example in the con text of to use the on tology ha v e a p ositiv e attitude to w ards
h uman resource managemen t if in some compan y only it Dictating the use of a particular on tology to p eo
senior managers can lead pro jects and Mr P aton is ple to whic h they ha v e not con tributed is not lik ely
pro ject leader then w e can deduce that Mr P aton is to succeed Preferably an on tology is constructed in a
V Benjamins D F ensel A G omez P er ez
Concept Component Concept Cylinder
Subclassf Component
Relation Partf Partf Engine
Number of arguments
Type of argument component Concept Crankshaft
Type of argument component Subclassf Component
Partf Engine
Concept Engine
Figure P art of a ph ysical device on tology
Subclassf Component

collab orativ e ert of domain exp erts represen tativ es
Figure P art of a mec hanical device on tology
of end users and IT sp ecialists Suc h a join t ert re
quires the use of a metho dology that guides the
on tology dev elopmen t pro cess and to ols to insp ect

tml
bro wse co dify mo dify and do wnoad the on tology
eadTITLE Mr Paton TITLE
Examples of suc h metho dologies include Methon
ONTOpagerojectLeader a
tology GJ GP Usc hold and Gruninger
head
metho dology G and that of Gruninger and F o x
ody

F The to ol w e use is the On tology Serv er
ONTOpageastNameody Pato n a
FR whic h is an in teractiv e en vironmen t esp e

cially useful for up dating main taining and bro wser
body
on tologies On tolingua on tologies can b e translated html

to diren t languages including Prolog CORBA
IDL HE CLIPS LOOM ac KIF Epikit
Figure A simple extension to HTML The onto
en On tologies built in On tolingua use the F rame
attribute allo ws to express on tological information in
On tology ru whic h is written in KIF no wl
HTML pages
edge In terc hange F ormat F The F rame On tol
ogy is as its name suggests a frameased language
Characterizing the kno wledge
whic h includes primitiv es suc h as classes sublasses
attributes v alues relations and axioms Related on As already men tioned brie in our approac h the
tologies can b e connected to eac h other b y inclusion kno wledge to b e managed is distributiv ely organized in
HTML pages in a compan y in tranet or on the
As an example consider the con text of the automo
WWW The relev an t kno wledge can th us b e main
bile industry Here the on tology w ould include among
tained distributiv ely b y diren t p ersons he resp on
others terms related to mec hanical and h ydraulic de
sible p ersons for the resp ectiv e HTML pages The
vices In the mec hanical device on tology examples of
sub ject matter kno wledge within the HTML pages is
classes are ylinder rankshaft and ngine An
annotated using the on tology as a sc heme for express
example of a binary relation is artf whic h could
ing metaata F or example in the h uman resource
b e used to sa y that the cylinder is partf the engine
managemen t domain the homepage of Mr P aton
The h ydraulic device on tology could include the class
w ould state that he is a pro ject leader W e th us add
ip e and the ternary relation onnection to ex
metaata to mak e this explicit In our approac h w e
press that t w o mec hanical devices are connected b y a
do this b y extending HTML with a new attribute of
giv en kind of pip e Note that the terms ylinder
the nc hor tag the onto attribute Figure giv es
rankshaft and ngine will b e part of an on tol
a simple illustration
ogy in the domain of ec hanical devices while the
The HTML co de in Figure states that the URL
concept omp onen t and the relation artf will
of the page con taining the information represen ts
b elong to a metan tology applicable to an y kind of
a ProjectLeader term deed in the on tology
ph ysical device Figure and Figure illustrate re
Page in h a ONTOpagerojec tL ea der i refers to
sp ectiv ely part of a ph ysical device on tology and part
the URL of the w eb page Body refers to what fol
of a mec hanical device on tology
lo ws and what is within the scop e of the anc hor i
In a h uman resource managemen t con text classes un til the closing h i The onto attribute do es not af
could b e mplo y ee anager ro ject leader fect the visualization of HTML do cumen ts in standard
kill rea of exp ertise Applied to a concrete com w eb bro wsers suc h as Netscap e or Explorer The only
pan y an on tology can full the role of an n terprise thing that it do es is that it mak es visible the sub ject
kno wledge map matter kno wledge for the in telligen t w eb cra wler This
V Benjamins D F ensel A G omez P er ez
small extension of HTML has b een c hosen to k eep an A W e describ e th us a virtual organization con
notation as simple as p ossible Also it enables the sisting of researc hers univ ersities pro jects publica
direct usage ctually reuse of textual kno wledge al tions etc The information resides at the W orld
ready in the b o dy of the anc hor This prev en ts the Wide W eb in the homepages of the KA researc hers
kno wledge annotater from represen ting the same piece where they publish information ab out their aiation
of information again he text Paton app earing as the pro jects publications researc h in terests etc F
v alue of metaata onto ab o v e is the same text as is
F rom a concrete kno wledge managemen t p oin t of

visualized in the bro wser This simple solution suf
view the A initiativ e is not an esoteric aca
es for our approac h b ecause the HTML pages only demic to y example Imagine a large m ultinational
con tain factual kno wledge DES with thousands of emplo y ees w orld wide F or suc h a
large organization ectiv e h uman resource manage
men t RM is of vital imp ortance Ho w ev er ding
In telligen t kno wledge retriev al
ho kno ws what in large organizations has alw a ys
Ha ving discussed the on tology and the annotated
b een a timen tensiv e pro cess A kno wledge manage
HTML pages w e will no w turn to using this kno wledge
men t system that allo ws to d adequate p eople based
for in telligen t retriev al W e use the on tologyased
on their skills exp erience and area of exp ertise w ould

brok ering service On tobrok er whic h consists of three
certainly b e of high v alue F or large companies that
main elemen ts a w eb cra wler alled On to cra wler an
ha v e an organizationide in tranet our approac h is a
inference engine and a query in terface DES
real p ossibilit y to enhance the HRM task It allo ws
First On to cra wler searc hes through the annotated
impro v emen t of the precision recall and presen tation
pages on an in tranet and collects the annotated
of the results of searc hes on an in tranet or the WWW

kno wledge fragmen ts Second it translates the anno
Notice ho w ev er that the fact that A is nat
tated kno wledge fragmen ts in to facts form ulated in the
urally related to the HRM task do es not imply that
represen tation language used b y On tobrok er Neither
it is limited to this kno wledge managemen t task In
the inference engine nor the querying user ha v e to b e
principle the sub ject matter of our approac h can con
a w are of the syn tactical w a y in whic h the facts are rep
cern an y kind of compan yital kno wledge that need
resen ted on the In ternet Only the annotaters ha v e to
to b e managed more ectiv ely
use the annotation language

The inference engine receiv es the query of a user and
On tological engineering in A
exploits t w o information sources for deriving an an

In A w e build an on tology of the KA comm unit y
sw er the on tology of the sub ject matter and the facts
f an n terprise kno wledge map Since an on tol
that w ere found b y On to cra wler The basic inference

ogy should capture consensual kno wledge in A
mec hanism of the inference engine is the deriv ation of a
sev eral researc hers co op erate together at diren t
minimal mo del of a set of Horn clauses ee DES
lo cations to construct the on tology In this w a y w e
for details This resem bles in telligen t reasoning as
ensure that the on tology will b e accepted b y a ma jorit y
kno wn in Kno wledgeased Systems with the dir
of KA researc hers The curren t on tology for the KA
ence that the instances of the kno wledge base are no w
comm unit y consists of sev en related on tologies an or
distributed o v er the diren t HTML pages The query
ganization on tology a pro ject on tology a p erson on tol
in terface of On tobrok er consists of a h yp erb olic visual
ogy a researc hopic on tology a publication on tology
ization of the on tology and a table format in whic h the
an ev en t on tology and a researc hro duct on tology
user can easily comp ose queries ee Figure This
The curren t v ersion of the on tology can b e view ed at
prev en ts the user from ha ving to kno w all the classes
the Europ ean mirror site in Madrid of the On tology
and attributes of the on tology

Serv er of Stanford Univ ersit y Login as n tologias
k a with passw ord dieu and then load one of

Pro of of concept A
the sev en subn tologies of the KA comm unit y F or
illustration purp ose w e include here examples of t w o
In order to in v estigate the feasibilit y of our approac h
subn tologies of the KA on tology the p erson on tol
w e are p erforming a largecale initiativ e on the W eb
ogy and the publication on tology
where the sub ject matter is the scien ti kno wledge
The P ersonn tology dees the t yp es of p ersons
acquisition comm unit y the Kno wledge Annotation
w orking in academic en vironmen ts along with their

Initiativ e of the Kno wledge Acquisition Comm unit y
c haracteristics This on tology dees classes and

relations The o v erview do es not sho w whic h classes
The URL of On tobrok er is
h ttpwwifbni arlsruheeBSrok er the relations connect ut it can b e bro wsed at On tol

The homepage of A is

h ttpwwifbni arlsruheeBSrok erAh tml URL is h ttpwwslv ciaiapms
V Benjamins D F ensel A G omez P er ez ogy Serv er Inden tation denotes the sub classf rela an annotated homepage of a researc her using the onto
tion attribute Page in h a ONTOpageddre ss bo dy i
refers to the URL of the w eb page Body refers to what
Class hierarchy classes defined
follo ws and what is within the scop e of the anc hor i
un til the closing h i Address is a class of the KA
Person
on tology Figure illustrates the annotation of a pub
Employee
Academictaff
lication The annotation pro cess lo oks lik e a tedious
Lecturer
and errorrone task Our exp erience is that it tak es
Researcher
roughly one hour to annotate e pages A t the On
Administrativetaff
Secretary tobrok er site an annotation c hec k er is a v ailable and
Technicaltaff
if needed p ersonal supp ort can b e giv en In spite of
Student
the amoun t of w ork in v olv ed there is one imp ortan t
Phdtudent
factor that ma y mak e p eople b e willing to annotate
relations defined
their homepages and that is selfublicit y By anno
tating pages researc hers mak e themselv es more visible
Address Affiliation Cooperatesith Editorf
to others whic h enhances the lik eliho o d that others
Email Firstame Hasublication Headfroup
will use and refer to their w ork whic h in the aca
Headfroject Lastame Memberfrganization
Memberfrogramommitt ee Memberfesearchroup
demic w orld is a go o d thing In Section w e come
Middlenitial Organizerfhairf Personame
bac k to this issue
Photo Researchnterest Secretaryf Studiest
Supervises Supervisor Workstroject
Querying the KA comm unit y
The Publicationn tology dees in classes and

In A in order for On to cra wler to collect the
relations the usual bibliographic en tities and at
kno wledge from HTML pages researc hers ha v e to reg
tributes
ister their pages That is they ha v e to tell On to cra wler
Class hierarchy classes defined
whic h URLs it needs to visit Once that is done in tel
ligen t kno wledge retriev al is p ossible Users are freed
Onineublication
from kno wing the sp eci querying language through a
Publication
Article
user in terface comprising a h yp erb olic visualization of
Articlenook
the on tology link ed with a table in terface ee Figure
Conferenceaper
and Figure In the h yp erb olic view the on tology
Journalrticle
Technicaleport can b e mo v ed around with the ect that concepts
Workshopaper
dragged to the cen ter are enlarged while p eripheral
Book
concepts are reduced in size If the user clic ks on a
Journal
concept it is passed to the table in Figure Sp eci
IEEExpert
attributes of the selected concepts can no w b e c ho
IJHCS
Specialssue
sen uc h as astname and mail In this w a y
users can comp ose their query b y bro wsing and clic k
relations defined
ing with a minim um amoun t of t yping The table also
Abstract Bookditor allo ws the construction of comp osite queries using con
Conferenceroceedingsi tle
junctiv es suc h as and or and not or not
Containsrticlenook
W e can for instance ask for all researc hers in the
Containsrticlenourn al Describesroject
KA comm unit y The answ er w ould not only include
Firstage Hasuthor Hasublisher Inook
Inonference Inournal Inrganization
researc hers who ha v e their homepage annotated but
Inorkshop Journalditor Journalumber
also additional researc hers that co op erate with these
Journalublisher Journalear Lastage
researc hers The on tology dees co op eration b et w een
Onineersion Onineersionf
Publicationitle Publicationear researc hers whic h enables the follo wing deduction
Technicaleportumber Technicaleporteries
if X co op erates with Y then X and Y m ust b e re
Type Volume Workshoproceedingsitl e
searc hers On tobrok er uses this t yp e information not
for consistency c hec king hic h w ould not b e a v ery

Annotating pages in A
go o d idea in an op en w eb en vironmen t but for ab duc

Annotating HTML pages in A means that eac h tiv ely deriving new facts Y is also a researc her
participating researc her in the KA comm unit y has to This example illustrates that it is p ossible to access
annotate the relev an t kno wledge in his or her home kno wledge that is not explicitly represen ted whic h is
page en vironmen t Figure illustrates fragmen ts of an imp ortan t adv an tage of our approac h compared to
V Benjamins D F ensel A G omez P er ez
tml
eadTITLE Richard Benjamins TITLE
ONTOpageesearcher a
head
HREFpicturesdich if
MG aligniddle SRCpicturesichard gif A
ONTOpagehotoref
HREFhttpwwiia sic es ic har di ctu res ric hard i f
ONTOpageirstNameod y ic har da
ONTOpageastNameody Benj ami ns a
h
ONTOpageffiliation ody HREFard
Artificial Intelligence Research Institute IIA
hrefhttpwwsics CSIC a Barcelona Spain r
and r
ONTOpageffiliation ody HREFhttpwwwi sy uva nl
Dept of Social Science Informatics WI

HREFhttpwwval uva ng lis h UvA A Amsterdam the
Netherlands
L
TSTRONGA HREFIIIAtml II A A
ONTOpageddressody
Artificial Intelligence Research Institute STRONG
TEMSIC Spanish Scientific Research CouncilEM
Tampus UAB
T Bellaterra Barcelona Spain a
TIMG SRCgifselif
voice
TIMG SRCgifsaxif
fax
TIMG SRCgifsmailif
EmailA HREFmailtoichardii as ic es ONTOpagemailref
richardiiasicsA
TRL HREFhttpwwiia csi cs r icha rd
httpwwiiasics ric hard A
DLont
body
html

Figure Example w eb page annotated with the ONTO attribute Page in h a ONTOpageddre ss bo dy i
refers to the URL of the page Body refers to what follo ws and what is within the scop e of the anc hor i un til
the closing h i Address is a class of the KA on tology
V Benjamins D F ensel A G omez P er ez
nameBenjamins ONTOnameournalArticle a
nameBenjamins ONTOnameuthorref
hrefhttpwwiias ic s rich ard nd ex html
V R Benjamins a
and
nameBenjamins ONTOnameuthorody
M Aben a
nameBenjamins ONTOnameitleody
Structurereserving KBS Development through
Reusable Libraries a Casetudy in Diagnosis a
nameBenjamins ONTOnameournalody
IJHCS a
International Journal of Humanomputer Studies Vol pages
nameBenjamins ONTOnameirstPageody
a
nameBenjamins ONTOnameastPageody
a
nameBenjamins ONTOnameearody
a
nameBenjamins ONTOnamenlineVersion href
HREFhttpwwiia sic es ic har do sts crip ts ijhc s s
raft version a

Figure Example of an annotated publication All v alues of the ONTO attribute b elong to the on tology of the
kno wledge acquisition comm unit y The actual kno wledge he instances represen ting the publication app ears at
the leftand side the righ t part con tains the annotation co de
k eyw ordased searc h W e could also ask for all re and main taining the on tology annotating in
searc hers that ha v e w ork ed together in some pro ject formation sources and querying them ee Fig
or for abstracts of all pap ers on a particular topic ure Curren tly w e use ODE F GPGP n
More examples of queries to the kno wledge acquisi tological design en vironmen t whic h allo ws one
tion comm unit y can b e obtained through On tobrok er to sp ecify on tologies at the conceptual lev el b y
homepage completing tables rather than at the implemen
tation lev el F rom these tables ODE is able to
generate the On tolingua co de of the on tology W e
Some facts
need ho w ev er to complemen t this with more sup
The curren t v ersion uly of the on tology con
p ort F or instance W eb on to om enables
tains classes axioms and attributes whic h
collab orativ e construction of on tologies o v er the
are used to annotate facts of researc hers
WWW Concerning the annotation pro cess w e
w ould need a to ol that visualizes b oth the on tol
F easibilit y of kno wledge manage ogy and the HTML page to b e annotated Select
ing a fragmen t of the HTML page and then clic k
men t systems
ing on a term of the on tology should ha v e to ect
In order to sa y something ab out the feasibilit y of a
to include the corresp onding onto attribute alue
horizon tal kno wledge managemen t system suc h as w e
in the HTML page
ha v e describ ed w e ha v e to consider the risks in v olv ed
Similarly to ols are needed for up dating kno wl
Risks come from v arious resources and w e will discuss
edge b oth at the instance lev el where researc hers
them resourceise tec hnological risks and so cial and
annotate their p ersonal pages as w ell at the on
organizational risks
tology lev el Changes to the on tology migh t ha v e
dramatic consequences for up dating the annota
T ec hnological risks
tions in HTML pages esp ecially in pages that are
annotated with an on tology term that b ecomes
F rom a tec hnology p oin t of view there are sev eral fac
obsolete W e do not ha v e a crystallized answ er for
tors that endanger the success of our kno wledge man
this problem y et but it certainly forms a risk to
agemen t approac h
b e considered One p ossibilit y w ould b e to use so

called XML ame spaces that let y ou include
First of all suc h an initiativ e is lik ely to fail with
in a do cumen t hen an XML page rather than
out dedicated to ols to supp ort the tasks in v olv ed

In particular to ols are needed for constructing
URL of XML is
V Benjamins D F ensel A G omez P er ez Figure The h yp erb olic query in terface Clic king on a no de mak es the corresp onding class app ear in the table
in terface of Figure
Figure The table query in terface
V Benjamins D F ensel A G omez P er ez In man y companies the men talit y is comp etitiv e
Tool for maintaining
rather than collab orativ e In other w ords f m y
ontologies
colleague wins then I lo ose And f I mak e
m y kno wledge a v ailable to others then others will
manipulate
pro from that and there will b e a risk that they
outp erform me This men talit y is a real threat
to success of kno wledge managemen t initiativ es
Ontology Increasingly more companies b ecome a w are that
use
use
a collab orativ e men talit y leads to b etter results
than comp etitiv e thinking o v Organizations
Annotation and Tools for retrieving can stim ulate collab orativ e thinking b y c hanging
wrapper generation
information and the incen tiv e system uc h as making it ancially
tools
answering queries
rew arding to share kno wledge
Giv en the high w orkload of to da y emplo y ees it
ma y b e easily felt that con tributing to a kno wl
HTML pages
enrich
use
(information edge managemen t ert is a w aste of time or at
sources)
least do es not ha v e priorit y This is killing for an y
kno wledge managemen t initiativ e Organizations
Figure T o ols to supp ort kno wledge managemen t should therefore rew ard kno wledge managemen t
con tributions equally as results that lead to di
HTML where the deition of a terms comes
rect pros In addition an ert should b e made
from
to reuse existing do cumen ts suc h that kno wledge
What happ ens when the kno wledge is spread o v er
w ork ers do not ha v e the impression that they ha v e
ten thousands of HTML pages Apart from the
to duplicate kno wledge There exist already to ols
up dating problem ee ab o v e also the in telligen t
to generate HTML pages from a v ariet y of other
reasoning part migh t b ecome a problem This is
formats SW ord Email etc
a familiar problem in KBS researc h when algo
rithms dev elop ed and tested on to y domains ha v e
Discussion and conclusion
to scalep to real w orld applications
Ho w do es our simple extension to HTML relate to
Summary
new tec hnologies for the W eb that migh t mak e
In this article w e presen ted a kno wledge engineering
HTML obsolete The W the in ternational
approac h to kno wledge managemen t whic h is based on
Worldide Web Consortium for dev eloping and
man y y ears of exp erience in dealing with kno wledge
promoting standards for the W eb curren tly in
If w e relate our w ork to the four kno wledge manage
tro duces the eXtensible Markup Language ML
men t actions men tioned in the in tro duction w e get the
as a new standard for expressing the structure
follo wing
of w eb do cumen ts and the Resource Description

F ramew ork DF for describing the seman tics
Kno wledge gathering is p erformed from existing
of w eb do cumen ts When a al v ersion of RDF
HTML pages no wledge annotation
is recommended b y the W w e will implemen t
Kno wledge organization and structuring is done
a wrapp er that automatically generates RDF def
through an on tology n tological engineering
initions from our annotations DES
Kno wledge reemen t is p erformed distributiv ely
b y eac h w ork er p date annotations
Kno wledge distribution is done b y a w eb cra wler
So cial and organizational risks
that giv es in telligen t access to the kno wledge that

is anaged This is a pull approac h where
Without participating researc hers the A ini
users tak e the initiativ e when they need kno wl
tiativ e w ould certainly fail Ho w ev er the nature
edge Ho w ev er the w ork presen ted here could as
of the initiativ e is suc h that participation is re
w ell b e used for a push approac h
w arding It is a selfromoting activit y That is
researc hers are b etter of if they participate b e
A so cial ert
cause other researc hers and outsiders can b etter
and more easily d their w ork
W e noted that kno wledge managemen t essen tially in
v olv es p eople and therefore an y kno wledge manage
h ttpwworgML

men t ert is do omed to fail if h uman factors are not
URL of RDF is
h ttpwworgetadataDFroupDdfyn tax tak en seriously Kno wledge managemen t only w orks if
V Benjamins D F ensel A G omez P er ez p eople co op erate and are willing to share their kno wl o v erwhelming amoun t of answ ers eferences to w eb
edge One w a y to stim ulate sharing of kno wledge is to do cumen ts In other w ords there is an information
c hange the incen tiv e system accordingly o v erload whic h mak es it hard to d exactly
what one is lo oking for and to get rid of nonsense ith
The kno wledge c hain resp ect to the query Although searc h engines get in
creasingly smarter w e exp ect that there will b e a limit
An imp ortan t framew ork in kno wledge managemen t is
to suc h k eyw ordased information retriev al More
the soalled no wledge Chain ou whic h refers
o v er curren t k eyw ordased searc h approac hes do not
to the adaptabilit y of an organization to an ev er c hang
allo w to presen t information collected from distribu
ing mark et The kno wledge c hain consists of four
tiv e lo cations in a coheren t w a y to users since there is
stages whic h are w alk ed through in a circular w a y
no kno wledge of ho w the retriev ed information relates
to eac h other On tologyased retriev al do es allo w for
In ternal Aw areness refers to the organization
this through the on tology Finally the on tologyased
abilit y to understand itself in terms of the skills
approac h allo ws to access implicit kno wledge whic h is
and comp etencies that it p ossesses and not so
deitely b ey ond the capacit y of k eyw ordased ap
m uc h in terms of its pro ducts
proac hes
In ternal resp onsiv eness is concerned with the
T o reduce the annotation ert mac hine learn
translation of in ternal a w areness kills and com
ing tec hniques can b e used that exploit on tologies to
p etencies in to teams with the skills and to ols to

automatically classify textual information DF
bring a pro duct to mark et
Moreo v er wrapp ers can b e built that extract the se
External resp onsiv eness mak es the dirence for
the organization success or failure It is the or man tics of w eb do cumen ts based on regularities in
ganization abilit y to tak e quic k and adequate de their structure format and con ten t Again mac hine
cisions based on a corp orate instinct rather than learning tec hniques can b e used to semiutomatically
to go through a long bureaucratic pro cess b efore build suc h wrapp ers K KWD Clearly this is
acting an imp ortan t researc h line to em bark on
External a w areness represen ts an organization
abilit y to understand ho w the mark et p erceiv es Related w ork
the v alue asso ciated with its pro ducts and ser
There is a h uge researc h ert going on ab out meta
vices as w ell as the c hanging directions and re
data for w eb do cumen ts XML RDF W ebSQL
quiremen ts of its mark ets When coupled with
Dublin Core More recen tly there are also sev eral
in ternal a w areness external a w areness ma y lead
pro jects that use on tologies together with metaata to
to en tirely new mark ets
impro v e information retriev al SHOE On tology
Markup Language Conceptual Kno wledge Markup
Our approac h con tributes directly to the st t w o
Language Most of these pro jects relate in some w a y
stages in ternal a w areness and resp onsiv eness An

or another to our approac h and to A in particular
on tological engineering pro cess as is part of our ap
W e already men tioned that the Resource Description
proac h results in a kno wledge map of the organiza
F ramew ork DF ma y pro vide an alternativ e syn tax
tion This ap certainly con tributes to the in ternal
for writing on tological annotations of w eb do cumen ts
a w areness of an organization The annotation pro cess
Metaata deed in RDF ha v e to b e pro vided on an
pro vides all instances of the kno wledge map Concern
extra page or in a blo c inside a w eb page There
ing the organization in ternal resp onsiv eness if eac h
fore elemen ts of a w eb page suc h as text fragmen ts or
emplo y ee annotates his or her homepage with skills
links cannot directly b e annotated with seman tics but
comp etencies and areas of exp ertise it will b e easy to
m ust b e rep eated in order to b e enric hed with meta
d quic kly and accurately the righ t p ersons for form
information This design decision ma y cause prob
ing adequate teams
lems for main taining w eb do cumen ts due to the redun
dancy of the information See h ttpwwifbni
On tologyased v ersus k eyw ordased re
k arlsruheeBSrok ernhaltp tml for brief
triev al
o v erviews of these related pro jects and links to their
One could argue that if all the kno wledge is a v ail
homepages
able in HTML do cumen ts then wh y use an on tol
ogy to annotate the information in the pages Af
Ac kno wledgmen t
ter all the annotation ert is considerable Wh y not
use general searc h engines for k eyw ordased searc hing W e thank Stefan Dec k er for his tec hnical con tribution
through the HTML pages As ev eryb o dy migh t ha v e to On tobrok er Ric hard Benjamins is supp orted b y
exp erienced k eyw ordased searc h easily leads to an the Netherlands Computer Science Researc h F ounda
V Benjamins D F ensel A G omez P er ez tion with ancial supp ort from the Netherlands Or In Spring Symp osium Series on Ontolo gic al
Engine ering Stanford AAAI Press
ganization for Scien ti Researc h W O
en M R Genesereth editor The Epikit man
References
ual Epistmemics Inc P alo Alto CA
K N Ashish and C A Knoblo c k Semi
F M R Genesereth and R E Fik es Kno wl
automatic wrapp er generation for in ternet
edge in terc hange format v ersion refer
information sources In Pr o c e e dings of the
ence man ual T ec hnical rep ort Logic
IF CIS Confer enc e on Co op er ative Informa
Computer Science Dept Stanford Univ er
tion Systems o opIS Charlston South sit y h ttpwwsm b cduse
Carolina
F M Gruninger and M F o x Metho dology
PH J Angus J P atel and J Hart y Kno wledge
for the design and ev aluation of on tolo
managemen t Great concept but what is
gies In Pr o c e e dings of the Workshop on
it Information We ek
Basic Ontolo gic al Issues in Know le dge Shar
ing held in c onjunction with IJCAI Mon
F V R Benjamins and D F ensel The on

treal Canada
tological engineering initiativ e A In
N Guarino editor F ormal Ontolo gy in
P A G omez erez Kno wledge sharing and
Information Systems pages IOS
reuse In J Lieb o witz editor The Hand
Press
b o ok of Applie d Exp ert Systems CR C

F GPGP M Bl azquez M F ern andez J M Garc ia
ru T R Grub er A translation approac h to
Pinar and A G omez erez Building on
p ortable on tology sp eciations Know le dge
tologies at the kno wledge lev el using the
A c quisition
on tology design en vironmen t In Pr o c e e d
ua N Guarino F ormal on tology concep
ings of the Eleventh Workshop on Know l
e dge A c quisition Mo deling and Manage tual analysis and kno wledge represen tation
ment KA W Ban Canada International Journal of Humanomputer
Studies Sp ecial is
or W N Borst Construction of Engine er
sue on The Role of F ormal On tology in the
ing Ontolo gies PhD thesis Univ ersit y of
Information T ec hnology
Tw en te Ensc hede

ou Thomas Koulop oulos Ho oking in to the
DF M Cra v en D DiP asquo D F reitag A Mc
kno wledge c hain White pap er at KM
Callum T Mitc hell K Nigam and S Slat
W orldom
tery Learning to extract sym b olic kno wl
edge from the w orld wide w eb In Pr o c e e d
WD N Kushmeric k D W eld and R Do oren
ings of the th National Confer enc e on AI
b os W rapp er induction for information ex
AAI Madison Wisconsin
traction In Pr o c e e dings of the th Interna
tional Joint Confer enc e on AI JCAI
o v R Co v ey The seven habits of highly e c
tive p e ople Simon Sc h uster Inc New pages Nago y a Japan
Y ork
ac R MacGregor Inside the LOOM classir
om J Domingue T adzebao and w eb on to Dis
SIGAR T Bul letin June
cussing bro wsing and editing on tologies on
HE R Orfali D Hark ey and J Edw ards edi
the w eb In Pr o c e e dings of the Eleventh
tors The Essential Distribute d Obje cts Sur
Workshop on Know le dge A c quisition Mo d
vival Guide John Wiley Sons New Y ork
eling and Management KA W Ban

Canada
D Oeary The in ternet in tranets and the
DES D F ensel S Dec k er M Erdmann and
AI renaissance IEEE Computer
R Studer On tobrok er The v ery high idea

In Pr o c e e dings of the th International
Flairs Confer enc e LAIRS Sanibal Is
D Oeary Kno wledge managemen t T am
land Florida
ing the information b easts IEEE Intel ligent
Systems Sp ecial Issue
FR A F arquhar R Fik es and J Rice The on
with three con tributions
tolingua serv er a to ol for collab orativ e on
tology construction International Journal
AA A Th Sc hreib er J M Akk ermans A A
of Humanomputer Studies a y
Anjewierden R de Ho og N R Shadb olt

W V an de V elde and B J Wielinga En
GJ M F ernandez A G omez P erez and N Ju gine ering and Managing Know le dge The
risto METHONTOLOGY F rom on to CommonKADS metho dolo gy to app ear
logical art to w ards on tological engineering
V Benjamins D F ensel A G omez P er ez BF R Studer V R Benjamins and D F ensel
Kno wledge engineering principles and
metho ds Data and Know le dge Engine ering

G M Usc hold and M Gruninger On tolo
gies principles metho ds and applications
Know le dge Engine ering R eview

HvdSK G v an Heijst R v an de Sp ek and
E Kruizinga Organizing cop orate memo
ries In B R Gaines and M A Musen ed
itors Pr o c e e dings of the th Ban Know l
e dge A c quisition for Know le dgease d Sys
tems Workshop pages Alb erta
Canada SRDG Publications Univ er
sit y of Calgary
SW G v an Heijst A T Sc hreib er and B J
Wielinga Using explicit on tologies in
KBS dev elopmen t International Journal
of Humanomputer Studies

ii K M Wiig Know le dge Management The
c entr al management fo cus for intel ligent
acting or ganizations Sc hema Press Ltd
Arlington TX
V Benjamins D F ensel A G omez P er ez