KNO

WLEDGE DISCOVERY AND EXCHANGE

Towards a Web-based Application for Discovery

and Exchange of Revealed Knowledge

S. J. Overbeek

e-ofﬁce B.V., Duwboot 20, 3991 CD Houten, The Netherlands, EU

P. van Bommel, H. A. (Erik) Proper, D. B. B. Rijsenbrij

Institute for Computing and Information Sciences, Radboud University Nijmegen

Toernooiveld 1, 6525 ED Nijmegen, The Netherlands, EU

Keywords:

Knowledge discovery, knowledge exchange, knowledge market.

Abstract:

Web technologies enable the discovery and exchange of knowledge from many different locations using many

different channels. This implies that one is able to discover and exchange knowledge while using a PDA when

traveling by train for instance. A provisional Web-based application referred to as ‘

DEXAR:

Discovery and

eXchange of Revealed knowledge’ is therefore introduced to illustrate the possibilities of the Web in the process

of knowledge discovery and exchange. This is illustrated by an example from the medical domain. Before

focussing on this Web application however a better understanding of knowledge discovery and exchange is

needed to be able to determine what kind of Web-based support is desired and feasible. Thus, a knowledge

market paradigm and a knowledge discovery paradigm are discussed in detail.

1 INTRODUCTION

Knowledge discovery and exchange might be asso-

ciated directly to (Web) technologies such as search

engines, agent technology, mining tools, meta-data

standards, query languages, query protocols, etcetera.

Obviously, Web technologies can assist in discovering

and exchanging knowledge from many different loca-

tions such as at home, at the ofﬁce or at the university.

Web technologies can also assist in discovering and

exchanging knowledge using a variety of channels,

such as a desktop computer, a notebook and a PDA.

Before focussing on a possible Web-based application

however a better understanding of knowledge discov-

ery and exchange is needed to be able to determine

what kind of Web-based support is desired and feasi-

ble.

The latter can be acquired by using several ref-

erence models (in terms of a knowledge market par-

adigm and a knowledge discovery paradigm respec-

tively) which depict essential knowledge market and

knowledge discovery mechanisms on a conceptual

level. The actual knowledge discovery mechanism

described only focuses on the discovery of revealed

knowledge. This is knowledge which is indeed known

to an individual or the organization. The actual dis-

covery of concealed knowledge is not elaborated in

depth, because it will probably ﬁt more in a knowl-

edge mining paradigm than in the knowledge discov-

ery paradigm introduced in this paper.

Based on the reference models a provisional Web-

based application is elaborated, together with an ini-

tial user interface to assist the user in discovering and

exchanging revealed knowledge. The application is

used in a medical context, in that an assistant radiolo-

gist acquires medical knowledge utilizing knowledge

discovery and exchange mechanisms.

In section 2 the fundamentals of knowledge dis-

covery and exchange are elaborated, because it is nec-

essary to understand what is going to be discovered

and exchanged. A knowledge market paradigm is

then described in section 3, followed by a knowledge

discovery paradigm in section 4 which is a specializa-

tion of the knowledge market paradigm. After the the-

ory has been discussed, a provisional Web-based ap-

plication implements the knowledge market and dis-

covery paradigms into practice in section 5. Section 6

brieﬂy compares our reference models with other ap-

proaches in the ﬁeld and outlines the beneﬁts of our

approach compared to others. Section 7 concludes

this paper.

J. Overbeek S., van Bommel P., A. (Erik) Proper H. and B. B. Rijsenbrij D. (2007).

KNOWLEDGE DISCOVERY AND EXCHANGE - Towards a Web-based Application for Discovery and Exchange of Revealed Knowledge.

In Proceedings of the Third International Conference on Web Information Systems and Technologies - Web Interfaces and Applications, pages 26-34

DOI: 10.5220/0001264600260034

 SciTePress

2 FUNDAMENTALS OF

KNOWLEDGE DISCOVERY

AND EXCHANGE

Exploring the fundamentals of knowledge is neces-

sary to gain a better understanding of that what we

would like to discover and exchange. Before elab-

orating on the notion of knowledge, it is relevant to

mention that it can be regarded as ‘wrapped’ in in-

formation, whilst information is ‘carried’ by data (ex-

pressions in a symbol language) (Liang, 1994).

To determine possible knowledge types which can

be discovered and exchanged, implicit knowledge and

explicit knowledge can be elaborated at ﬁrst. (Nonaka

and Takeuchi, 1995) distinguish implicit knowledge

and explicit knowledge. Implicit knowledge com-

prises knowledge which is implicitly present in peo-

ple’s heads, such as skills which are difﬁcult to make

explicit. The way a physician makes a decision for

speciﬁc treatment is related to such skills. Implicit

knowledge is closely related to what is generally ex-

perienced as intuition. Explicit knowledge comprises

knowledge which can be expressed in terms of facts,

rules, speciﬁcations or textual descriptions. The dif-

ference between implicit and explicit knowledge is

relevant with regard to Web-based knowledge dis-

covery and exchange, because the discovery and ex-

change of implicit knowledge will require different

forms of Web-based support than the discovery and

exchange of explicit knowledge.

Besides discerning implicit and explicit knowl-

edge another relevant distinction can be made. Some-

times knowledge is present while one is not aware

of that knowledge. This varies from hidden skills of

workers (for an individual or for the organization),

via knowledge which is present in documents but not

properly indexed, to knowledge which is hidden in

undiscovered patterns in data collections (the basis for

data mining). (Hoppenbrouwers and Proper, 1999) al-

ready introduced this difference between revealed and

concealed knowledge. In contrast to implicit and ex-

plicit knowledge the knowledge status is considered

instead of the fundamental knowledge type.

The difference between implicit and explicit on

the one hand and revealed and concealed on the other

hand can be depicted in a 2x2 matrix as is shown in

ﬁgure 1. The following combinations are then possi-

ble:

• Implicit & concealed knowledge: e.g. competen-

cies or expertise of a worker unknown to the orga-

nization;

• Explicit & concealed knowledge: e.g. valuable

insights concealed in available data collections (to

Implicit Explicit

Revealed

Concealed

Unknown

competences

Known

competences

Unknown

patterns and

structures in data

Documented

knowledge

Figure 1: Four knowledge types, adapted from (Hoppen-

brouwers and Proper, 1999).

be discovered by data mining);

• Implicit & revealed knowledge: e.g. known ex-

pertise of a worker which can be appealed to;

• Explicit & revealed knowledge: e.g. best-practice

documentation, knowledge bases, scientiﬁc pa-

pers, etcetera.

The four knowledge types require different forms of

Web-based support, varying from support which is

aimed at expertise analysis, pro-active knowledge ex-

change, human resource related information systems,

planning systems, data mining and documentation

management. In the broad sense, Web-based support

can be applied to all knowledge types, nevertheless

Web-based support is usually applied more easily to

explicit knowledge than implicit knowledge.

At this point we can aim on the actual knowl-

edge discovery and exchange, which is illustrated

by the reference models together with a provisional

Web-based application from the medical domain. Be-

fore focussing on discovery and exchange of revealed

knowledge only (in a knowledge discovery para-

digm), a more general knowledge market paradigm

is discussed.

3 THE KNOWLEDGE MARKET

PARADIGM

Several knowledge types have been discussed up till

now. In practice the difference between concealed and

revealed knowledge is especially of importance. Re-

vealed knowledge can be localized (even when it is

implicit), but concealed knowledge can not be local-

ized (even when it has been made explicit in the past).

These knowledge types materialize in a knowl-

edge market paradigm as is depicted in ﬁgure 2.

3.1 Roles in the Knowledge Market

As in every market model the sketched knowledge

market contains the following roles: a broker role, a

KNOWLEDGE DISCOVERY AND EXCHANGE - Towards a Web-based Application for Discovery and Exchange of

Revealed Knowledge

Offering

Supplying

Acquiring

Brokering

Asset Asset

Characterization Characterization

Knowledge

intensive

tasks

Repositories

Mining

Development

Elicitation

BrokerSupplier Utilizer

“Offer” “Market” “Demand”

Transporter

Figure 2: A knowledge market paradigm.

supplier role, a transporter role and a utilizer role.

Formally, this set of roles can be represented as:

R O , {broker,supplier,transporter,utilizer} (1)

The four roles of the knowledge market paradigm

wish to achieve speciﬁc objectives within the market

and enable to abstract from the actors that will eventu-

ally enact the role. A role enactment is a speciﬁc ful-

ﬁlment of such a role by any eligible entity, expressed

by the function Enact : R E → R O, where R E is the

set of all role enactments within the domain of the

knowledge market. Given the role enactment e of a

role Enact(e), we can view the actor that speciﬁcally

enacts the role as a function Player : R E → AC. Here

AC represents the speciﬁc set of actors. Each role is

discussed brieﬂy in this section, also illustrated by a

practical example. The following text about role en-

actments incorporates ideas from (Gils et al., 2006) to

illustrate the essence of role enactments in the knowl-

edge market. Since an enactment indicates an actor

‘in a role’ we know that an actor and a role combina-

tion uniquely determines an enactment:

Player(e

) = Player(e

) ∧ Enact(e

) = Enact(e

)⇒e

(2)

For an enactment e ∈ R E the following notation is

introduced:

ha, ri , e such that Player(e) = a∧Enact(e) = r (3)

This can be illustrated by the following example. Let

an assistant radiologist denoted by a be an actor that

can play two roles. He either plays the role of type

broker denoted by r

, or the role of type utilizer de-

noted by r

. Both e

= ha, r

i and e

= ha, r

i are

enactments such that Player(e

) = a, Player(e

) = a,

Enact(e

) = r

and Enact(e

) = r

3.2 Tasks in the Knowledge Market

An actor in the knowledge market executes several

tasks while playing a role. The utilizer for instance

‘utilizes’ knowledge by executing knowledge inten-

sive tasks as is depicted in ﬁgure 2. Actors in the

knowledge market execute different tasks depending

on the role they play. A speciﬁc fulﬁlment of such a

task is expressed by Task : T I → T A, where T I is

the set of tasks which are executed by an actor (re-

ferred to as task instances) and T A is the set of tasks.

Given the task instantiation i of a task Task(i), we can

view the actor that is enacting a role and also speciﬁ-

cally executing a task as a function Exec : T I → R E.

The essence of task execution in the knowledge mar-

ket can be illustrated in the same way as in the role

enactment part of section 3.1. Since an execution de-

notes an enactment in a task we know that an enact-

ment and a task combination uniquely determines an

execution:

Exec(i

) = Exec(i

) ∧ Task(i

) = Task(i

)⇒i

(4)

For a task instance i ∈ T I the following notation is

introduced:

he,ti , i such that Exec(i) = e ∧ Task(i) = t (5)

This can also be illustrated by an example. Assume

that a radiologist enacts a ‘utilizer’ role. This en-

actment is denoted by e. While enacting the role he

can execute a task of type patient monitoring denoted

by t

and a task patient diagnosing denoted by t

Both i

= he,t

i and i

= he,t

i are executions such

that Exec(i

) = e, Exec(i

) = e, Task(i

) = t

and

Task(i

) = t

Now that roles and tasks in the knowledge market

have been discussed on a conceptual level, the func-

tion of the knowledge market in terms of supply and

demand is elaborated in the next section.

3.3 Supply and Demand

The ‘merchandize’ within the paradigm consists of

knowledge assets KA. These assets are tradeable

forms of revealed knowledge, which are transported

physically by the transporter. In the context of a Web-

based application, a transporter role can be enacted by

an actor equalling a Web protocol on the application

WEBIST 2007 - International Conference on Web Information Systems and Technologies

layer according to the OSI model (ISO/IEC, 1994),

such as FTP, GOPHER, HTTP, IRC or TELNET.

Knowledge assets are not necessarily explicit

knowledge assets, though. Implicit knowledge inside

people’s heads is also tradeable, because one can take

its implicit knowledge to a certain situation where that

implicit knowledge is wanted. This is what e.g. physi-

cians do when explaining a patient’s status to a col-

league. In that case the air functions as a transporter.

In the context of the Web however, a knowledge as-

set can be deﬁned as any entity that is accessible on

the Web that can provide knowledge to other entities

connected to the Web.

The goal of a supplier is to deliver knowledge,

which requires a ‘client’ who would like to utilize the

knowledge. This is only possible if the supplier is able

to make clear what is on offer, hence it is vital that

the knowledge is correctly characterized. This is not

always an easy job because terminology issues can

throw a spanner in the works. Poor characterizations

can inevitably lead to supplying irrelevant knowledge,

or omitting to supply relevant knowledge.

On the supply side of the knowledge market vari-

ous resources can be accessed: repositories, data col-

lections, mining, active knowledge development or

questioning experts (elicitation). A reliable supplier

offers revealed knowledge which is localizable and

available. It is possible to offer implicit knowledge,

e.g. by means of a reliable expert, as long as is as-

sured that the implicit knowledge can be applied by

the utilizer (albeit a certain competence).

The potential utilizer is searching for knowledge,

but does not know if that knowledge can be found.

Often a utilizer does not even know which knowl-

edge is necessary to fulﬁl the need. The knowledge

is concealed for the potential utilizer, but does cer-

tainly not have to be concealed for the knowledge

supplier. Characterization is key here, which matches

the knowledge demand with the knowledge to be sup-

plied. The broker plays a very important role in

matching supply and demand. It can be said that

knowledge discovery comes into play when potential

utilisers do not know beforehand which knowledge is

required to fulﬁl their needs.

Now that we have focussed on supply and de-

mand in the knowledge market, the actual exchange

of knowledge (albeit knowledge assets or knowledge

which is exchanged in a characterization process)

within such a market is elaborated in the next section.

3.4 Levels of Knowledge Exchange

In the knowledge market paradigm, two levels of

knowledge exchange can be discerned: instance level

knowledge and meta level knowledge:

KL , {instance−knowledge,meta−knowledge} (6)

What is exchanged between the supplier, transporter

and utilizer can be classiﬁed as instance level knowl-

edge. On the instance level, the actual knowledge as-

sets which are utilized by an actor who enacts the uti-

lizer role are intended. When knowledge exchange

on the instance level is concerned, the knowledge

assets are part of the knowledge output of the sup-

plier as well as the transporter. The assets are then

input for the utilizer. Input and output of knowl-

edge (on the instance level) can be represented as

In, Out : R E → ℘(KA), where In(e) determines the

input in terms of knowledge assets of an actor enact-

ing a role e. The function Out(e) then determines the

output. So knowledge exchange on the instance level

can be represented as:

x, y, z ∈ R E ∧ Enact(x) = supplier ∧

Enact(y) = transporter ∧

Enact(z) = utilizer ⇒

In(y) ∩ Out(x) 6=

0 ∨ In(z) ∩ Out(y) 6=

0 (7)

The intersection of knowledge input and output is

however considered as an empty set when only a sin-

gle actor is enacting a role:

∀

e∈R E

[In(e) ∩ Out(e) =

0 ∧ MIn(e) ∩ MOut(e) =

0] (8)

A meta level of knowledge exchange always con-

tains a formulation in terms of a question or a

query which reasons about knowledge which a uti-

lizer wants to receive. Meta level knowledge com-

prises the knowledge which is exchanged between the

utilizer, the broker and the supplier in the process

of matching supply and demand. Just that knowl-

edge which ﬂows between the aforementioned three

roles in the characterization process is intended here.

This knowledge is dubbed characterization knowl-

edge, which is represented by the set CK . Input

and output of knowledge (on the meta level) can be

represented as MIn, MOut : R E → ℘(CK ), where

MIn(e) determines the input in terms of characteri-

zation knowledge of an actor enacting a role e. The

function MOut(e) then determines the output. So

knowledge exchange on the meta level can be repre-

sented as:

x, y, z ∈ R E ∧ Enact(x) = utilizer ∧

Enact(y) = broker ∧

Enact(z) = supplier ⇒

MIn(y) ∩ MOut(x) 6=

0 ∧

MIn(x) ∩ MOut(y) 6=

0 ∨

MIn(z) ∩ MOut(y) 6=

0 ∧

MIn(y) ∩ MOut(z) 6=

0 (9)

KNOWLEDGE DISCOVERY AND EXCHANGE - Towards a Web-based Application for Discovery and Exchange of

Revealed Knowledge

In the following section, a scenario is described in

which the knowledge market paradigm materializes

as a whole.

3.5 A Knowledge Market Scenario in

the Medical Domain

The knowledge market paradigm comes to life when

it is illustrated by a practical problem from the med-

ical domain. Assume that an assistant radiologist de-

noted by actor a

requires knowledge about pneu-

monia when executing a knowledge intensive task.

This task is referred to as patient diagnosing de-

noted by t

while enacting the role of utilizer de-

noted by u. Formally, e

= ha

, ui is an enactment

such that Player(e

) = a

and Enact(e

) = u. Also,

= he

i is a task execution such that Exec(i

) = e

and Task(i

) = t

The assistant radiologist ﬁnds a senior radiologist

denoted by a

. This actor plays the role of broker de-

noted by b. Let e

= ha

, bi be an enactment such

that Player(e

) = a

and Enact(e

) = b. Actor a

explains his speciﬁc knowledge need expressed by

MIn(e

) ∩ MOut(e

) 6=

0. To fulﬁl the task assis-

tant support the senior radiologist probably asks some

additional questions to understand what the assistant

needs which is expressed by MIn(e

) ∩ MOut(e

) 6=

0. The broker’s task fulﬁlment can be formalized as

follows: i

= he

i is a task execution such that

Exec(i

) = e

and Task(i

) = t

After this initial conversation the senior radiolo-

gist decides to show an X-ray of human lungs to the

assistant which he ﬁnds in a wiki (Wiki, 2006). The

X-ray clearly shows symptoms of pneumonia in the

lungs together with a textual explanation. Here, the

Wikipedia Web site is an actor a

which plays the role

of knowledge supplier s. Formally, e

= ha

, si is an

enactment such that Player(e

) = a

and Enact(e

) =

s. The task supplying radiology wiki which is exe-

cuted by s implies that i

= he

i is a task execution

such that Exec(i

) = e

and Task(i

) = t

Characterization between the senior radiologist

and Wikipedia can be illustrated by the events in

which the senior radiologist searches for usable text

and images until the returned results are sufﬁcient

to reduce the utilizer’s knowledge need. This spe-

ciﬁc characterization process can be formalized as

MIn(e

) ∩ MOut(e

) 6=

0 ∨ MIn(e

) ∩ MOut(e

) 6=

Eventually, the HTTP protocol (as part of the In-

ternet in essence) is an actor a

which enacts the role

of knowledge transporter p and transports the assets

to the utilizer. Hence, the task transporting radiol-

ogy wiki is executed by the transporter. Therefore,

= ha

, pi is an enactment such that Player(e

) = a

and Enact(e

) = p. Also, i

= he

i is a task exe-

cution such that Exec(i

) = e

and Task(i

) = t

. The

exchange of knowledge assets can be expressed by

In(e

) ∩ Out(e

) 6=

0 ∨ In(e

) ∩ Out(e

) 6=

A wiki is an example of explicit knowledge. A

typical situation of a knowledge market involving im-

plicit knowledge can also be found in the medical do-

main. Assume that a physician working at the radi-

ology department of a hospital has a relatively good

overview of his colleagues’ expertise. This physician

functions as a broker in the knowledge market. As-

sume that a colleague has a question about tuberculo-

sis symptoms so that he becomes a potential knowl-

edge utilizer. The local knowledge broker can re-

fer him to another physician who knows more about

those symptoms. This colleague will then become

a potential knowledge supplier. If the utilizer starts

a conversation with the supplier the air which trans-

ports the sound of the words functions as a knowledge

transporter. If e-mail is used to ask a question then the

e-mail system will be the knowledge transporter.

In the next section the discovery and exchange of

revealed knowledge is discussed, using a specializa-

tion of the knowledge market paradigm.

4 DISCOVERY AND EXCHANGE

OF REVEALED KNOWLEDGE

Knowledge discovery of revealed knowledge is com-

plementary to information retrieval. The match-

ing and transportation of the data which carries this

knowledge is a controllable technical issue if every

party exactly knows which knowledge is required and

offered. Information retrieval is related with the au-

tomated or manual search for revealed knowledge

which is represented by an explicit characterization.

It is certain that the requested data (the carrier of the

sought knowledge) is not available if there is no match

and the used information retrieval mechanism is prop-

erly constructed. However, one may still ponder if

one has searched for the right knowledge. In other

words: has the query correctly characterized the need

for knowledge?

The knowledge discovery paradigm of ﬁgure 3 il-

lustrates the matter mentioned above. This paradigm

is based on the information discovery paradigm as

found in (Proper, 1999). This paradigm is a special-

ization of the knowledge market paradigm, aimed at

the discovery of revealed knowledge. Figure 3 shows

a trajectory with the knowledge gap of a utilizer as

starting point, or in other words the moment when

someone experiences a knowledge gap and the neces-

sity to ﬁll that gap. This leads to a need for knowl-

WEBIST 2007 - International Conference on Web Information Systems and Technologies

Knowledge gap

Knowledge need

Realize

Formulate

Formulation

Retrieve

Study

Utilizer

Brokering

Characterization

Results

Characterization

knowledge demand

knowledge supply

Supplier

Assets

Broker

Figure 3: A knowledge discovery paradigm.

edge. It is assumed that a need for knowledge is inﬂu-

enced by what the utilizer already has received from

the transporter in terms of assets. As introduced ear-

lier by (Weide and Bommel, 2006), this can be mod-

elled as a function:

Need : ℘(KA) × KA 7→ [0, 1] (10)

Need(S, a) is interpreted as the residual need for

knowledge asset a after the set S has been presented

to the utilizer, where S ⊆ KA. The set S can be in-

terpreted as the personal knowledge of a utilizer (also

called a knowledge proﬁle) during the discovery and

exchange of knowledge. No more knowledge is re-

quired by the utilizer if his need for knowledge deteri-

orates, which is denoted by Need(S, a) = 0. To lessen

a knowledge need the characterization of a knowledge

gap is necessary. This comprises the formulation of a

description of this knowledge need in terms of a ques-

tion (which contains the description but which is not

directed to someone), or a query (which communi-

cates the question to a machine). In the knowledge

discovery paradigm this is described as the knowledge

demand. In the ideal case a knowledge demand is for-

mulated by the utilizer and the broker. In a medical

context this can be compared to a conversation be-

tween an assistant radiologist (who requires speciﬁc

medical knowledge about a disease) and a senior radi-

ologist. The broker can support the utilizer in formu-

lating the knowledge demand which is a crucial step

in knowledge discovery (of revealed knowledge).

The supplier and the deliverable knowledge as-

sets are positioned opposite from the utilizer. Ideally,

these assets are described by means of a cooperation

between the supplier and the broker in terms of a char-

acterization. The broker can assure that this character-

ization is also used by the utilizer of this knowledge

while formulating their knowledge demand. The bro-

kering activities which are carried out by the broker

are in principle not different from an information re-

trieval process.

In order to have a graphical representation of the

deﬁnitions discussed in both the knowledge market

and discovery paradigms, an object-role model is pre-

sented in ﬁgure 4. For details on object-role models,

see e.g. (Halpin, 2001). Figure 4 depicts a special-

Characterization

Knowledge

Exchange

Level

(name)

Knowledge

Asset

Enactment

Actor

(id)

Role

(name)

Need

{‘instance-knowledge’,‘meta-knowledge’}

{‘broker’,‘supplier’,

‘utilizer’,‘transporter’}

0..1

Value

(number)

Execution

Task

(name)

Figure 4: Object-role model of knowledge market and dis-

covery paradigms.

ization relation between the subtypes Characteriza-

tion Knowledge and Knowledge Asset and a super-

type Knowledge. This implies that characterization

knowledge and each knowledge asset is, of course,

knowledge. For proper specialization, it is required

KNOWLEDGE DISCOVERY AND EXCHANGE - Towards a Web-based Application for Discovery and Exchange of

Revealed Knowledge

that subtypes are deﬁned in terms of one or more of

their supertypes. Such a decision criterion is referred

to as a subtype deﬁning rule. In ﬁgure 4 the subtype

deﬁning rule for Characterization Knowledge is:

CharacterizationKnowledge = Knowledge has

ExchangeLevel: ‘meta-knowledge’

The subtype deﬁning rule for Knowledge Asset is:

KnowledgeAsset = Knowledge has Ex-

changeLevel: ‘instance-knowledge’

The language used in the deﬁnitions of the subtype

deﬁning rules is described in e.g. (Hofstede et al.,

1993).

Thus far we have focussed on a knowledge market

paradigm, a scenario of a knowledge market in prac-

tice and a knowledge discovery paradigm. In the next

section a provisional Web-based application is intro-

duced to deliver support for activities within those

paradigms.

5 DEXAR: A WEB-BASED

APPLICATION FOR

DISCOVERY AND EXCHANGE

OF REVEALED KNOWLEDGE

In this section we discuss a way of support for bro-

kering and transporting activities as mentioned in the

knowledge market paradigm and partly in the knowl-

edge discovery paradigm, in the form of a provisional

Web-based application referred to as

DEXAR

: Discov-

ery and eXchange of Revealed knowledge. Further-

more, we introduce an initial user interface for this

application which illustrates interaction between a

physician and the application.

The

DEXAR

application implements a support

mechanism so that it keeps track of a utilizer’s knowl-

edge proﬁle by collecting a utilizer’s knowledge ques-

tions together with the knowledge supplied. Feed-

back of the utilizer to

DEXAR

creates insight in a uti-

lizer’s knowledge need. When a potential utilizer

wishes to acquire knowledge he can start to interact

with

DEXAR

. Assume that John Doe is an assistant ra-

diologist who would like to know more about pneu-

monia while studying at home. John opens his Web

browser and starts a conversation with

DEXAR

, which

is shown in ﬁgure 5. This is in principle a characteri-

zation process in which the application tries to charac-

terize the knowledge demand. After this conversation,

the broker application ﬁnds an image and text on the

Web which might be relevant for John Doe. Figure 6

shows how the application presents part of the results

to John. To determine if John still requires additional

http://DEXAR

Demand

Supply

Feedback

Profile

User: John Doe

Function: Assistant radiologist

John: “How can I identify a patient

suffering from pneumonia?”

DEXAR: “Would you like to use

radiology to identify the disease?”

John: “Yes.”

DEXAR: “Would you like to identify

a patient suffering from

a specific form of pneumonia?”

John: “Yes, I am interested

in identification of Q fever

DEXAR: “Okay, please wait while I will

try to find knowledge for you.”

pneumonia.”

Figure 5: Conversation with

DEXAR

http://DEXAR

Supply

Demand

Feedback

Profile

User: John Doe

Function: Assistant radiologist

DEXAR: “I have found an image along with

some explanatory text.”

John: “Okay, can I take a look at that image?”

DEXAR: “Of course, take a look at it below.”

Figure 6: Retrieving knowledge assets with

DEXAR

knowledge, which is the case if Need(S, x) > 0, a

feedback screen is shown which is depicted in ﬁg-

ure 7. John’s ﬁnal input in the conversation on the

screen of ﬁgure 7 indicates that Need(S , x) = 0. John

Doe can view his knowledge proﬁle which is shown in

ﬁgure 8. A history is generated showing John Doe’s

requests for knowledge and the results provided by

DEXAR

. Underlined words such as the word ‘image’

represent hyperlinks to the underlying Web pages.

So clicking on the word ‘image’ retrieves that image

from the Web. Furthermore, a lattice is constructed

containing the terms which the application has dis-

tilled from conversations with the user. The user may

browse through the lattice to learn about previously

recorded knowledge and to gain insight in the user’s

WEBIST 2007 - International Conference on Web Information Systems and Technologies

http://DEXAR

Feedback

Demand

Supply

Profile

User: John Doe

Function: Assistant radiologist

DEXAR: “Are you satisfied with the

knowledge provided?”

John: “Not exactly, I would like to view the

DEXAR: “Q Fever is a zoonosis caused by

the strictly intracellular, gram

negative bacterium Coxiella burnetii”

explanatory text too.”

John: “Thanks, that is enough for know.”

Figure 7: Feedback process in

DEXAR

http://DEXAR

Profile

Demand

Feedback

Supply

User: John Doe

Function: Assistant radiologist

17/06/2006 – Requested knowledge about

Q fever pneumonia.

17/06/2006 – Retrieved image

and text from

Wikipedia

History of knowledge discovery

John Doe’s (partial) knowledge profile

Q fever

pneumonia

zoonosis

identification

of a patient

pneumonia

identification of a patient

with Q fever pneumonia

identification of a patient

zoonosis

Q fever pneumonia zoonosis

with Q fever pneumonia

Q fever

Figure 8: Showing a (partial) knowledge proﬁle.

own proﬁle as a whole. In ﬁgure 8, John Doe’s knowl-

edge proﬁle is only partially displayed as a lattice.

The lattice in

DEXAR

is constructed by using index ex-

pressions as can be found in e.g. (Bruza, 1990). Index

expressions have the following syntax:

IdxExpr → Term{Connector IdxExpr}

∗

Term → String

Connector → String

The lattice shown in ﬁgure 8 resembles a partially dis-

played power index expression. A power index ex-

pression is the set of all index expressions, including

the empty index expression and the most meaningful

index expression. An example of an index expression

is (identiﬁcation of a patient) with (Q fever pneumo-

nia). Simply put, (power)index expressions are used

DEXAR

as a representation for a knowledge proﬁle.

Searching through a user’s own knowledge proﬁle

can be implemented by using Query by Navigation as

is described in e.g. (Grootjen, 2000). At ﬁrst, the user

may provide the application with an index expression

(in its shortest form this is a single Term). Once the

user is done specifying such a query the application

‘knows’ which knowledge discovery history (which

is coupled to the index expression) can be shown to-

gether with hyperlinks to relevant resources on the

Web.

Now that theoretical models and a possible way

of support for discovery and exchange of revealed

knowledge has been discussed, it is appropriate to

compare our approach with other approaches in the

ﬁeld. The next section therefore deals with this mat-

ter.

6 DISCUSSION

When studying the literature on theory and appli-

cations in the area of knowledge discovery and ex-

change, it is obvious that knowledge discovery, what-

ever the reason, is often equalled to data mining. Con-

sider for example (Roddick et al., 2003). When com-

pared with the speciﬁc distinction made in ﬁgure 1,

this means that when equalling knowledge discovery

with data mining one is focussing on explicit & con-

cealed knowledge. We believe that when discussing

the status and the type of knowledge one has more

comprehension of what can be discovered and ex-

changed than when solely understanding discovery of

knowledge to be data mining. Using the frame of

thought as depicted in ﬁgure 1 one can probably fo-

cus more easily on the discovery of one or more of the

knowledge types shown in the four quadrants.

In (Desouza and Awazu, 2003) an internal knowl-

edge market is deﬁned as a collection of buyers and

sellers who interact to determine the price of a product

or a set of products. The main components of an in-

ternal knowledge market are: the players (buyers and

sellers), rules (governance of interactions), and space

(area where buyers and sellers collect). Compared to

the knowledge market paradigm discussed in this pa-

per, the buyers and sellers are in accordance with the

utilizer and the supplier roles respectively. However,

there are no components that can be compared to the

transporter and the broker roles. To be able to focus

on the exchange of knowledge within a knowledge

market (which is an evident part of the research re-

ported in this paper) analysis of the interaction aimed

at delivery and broadcasting of knowledge between

buyers and sellers is necessary. An advantage of our

model is that those interactions are made clearer by

the introduction of additional roles and characteriza-

tion knowledge and knowledge assets. In our view,

a space component is not actually a component of a

KNOWLEDGE DISCOVERY AND EXCHANGE - Towards a Web-based Application for Discovery and Exchange of

Revealed Knowledge

knowledge market itself but is dependent of a speciﬁc

instantiation of a knowledge market. I.e. if roles in a

knowledge market are executed by actors who are lo-

cated at a (physical) library then the knowledge mar-

ket is part of a ‘physical’ space. If a knowledge mar-

ket mechanism as proposed in this paper takes part

online, other non-physical actors may be involved.

The rules shown in the model of (Desouza and

Awazu, 2003) however do intend to cover exchange

mechanisms within a knowledge market. These ex-

change mechanisms in a market should address what

goods will be bought and sold and how they will

be paid for. The pricing of knowledge is an impor-

tant issue to address, but it seems that the approaches

of (Desouza and Awazu, 2003) and also (Brydon and

Vining, 2006) are more identical to how traditional

economic markets function by primarily focussing on

price and volume of knowledge. Our models elab-

orate on the discovery and exchange of knowledge

more comprehensively instead because we speciﬁ-

cally address those topics and also believe that they

are a crucial part of knowledge market mechanisms.

7 CONCLUSION

This paper describes a vision on knowledge discov-

ery and exchange from a conceptual level, illustrated

by several reference models. Proceeding from these

models a Web-based application illustrates how the

reference models can be materialized within the med-

ical domain.

Future research is aimed at the possible applica-

tion of

DEXAR

at the radiology department of the Ni-

jmegen University Medical Centre St. Radboud, The

Netherlands, EU so that possible positive and negative

experiences made with

DEXAR

can be understood. A

goal is to support radiology students fulﬁlling knowl-

edge intensive tasks. This support is divided in two

parts: improving the students’ learning process and

improving the eventual product of learning, e.g. prac-

tical work and exams. Another possible application

DEXAR

in the area of information & knowledge sys-

tems modelling will be studied. The speciﬁc focus

in that area is to support modellers in the process of

modelling architecture principles. This should even-

tually lead to an improved product i.e. an improved

architecture.

REFERENCES

Bruza, P. (1990). Hyperindices: A novel aid for search-

ing in hypermedia. In Rizk, A., Streitz, N., and An-

dre, J., editors, Hypertext: Concepts, Systems and Ap-

plications; Proceedings of the European Conference

on Hypertext - ECHT 90, number 5 in Cambridge Se-

ries on Electronic Publishing, pages 109–122. INRIA,

Paris, France, EU, Cambridge University Press, Cam-

bridge, UK, EU.

Brydon, M. and Vining, A. (2006). Understanding the fail-

ure of internal knowledge markets: A framework for

diagnosis and improvement. Information & Manage-

ment, 43(8):964–974.

Desouza, K. and Awazu, Y. (2003). Constructing inter-

nal knowledge markets: Considerations from mini

cases. International Journal of Information Manage-

ment, 23(4):345–353.

Gils, B. v., Bommel, P. v., Proper, H., and Weide, T. v. d.

(2006). Quality makes the information market. In

Proceedings of the 14th International Conference on

Cooperative Information Systems (CoopIS), volume

4275 of Lecture Notes in Computer Science, pages

345–359. Springer-Verlag, Berlin, EU.

Grootjen, F. (2000). Employing semantical issues in syn-

tactical navigation. In Proceedings of the 22nd BCS-

IRSG Colloquium on IR Research, pages 22–33, Sid-

ney Sussex College, Cambridge, UK, EU.

Halpin, T. (2001). Information Modeling and Relational

Databases, from Conceptual Analysis to Logical De-

sign. Morgan Kaufmann, San Mateo, CA, USA.

Hofstede, A. t., Proper, H., and Weide, T. v. d. (1993). For-

mal deﬁnition of a conceptual language for the de-

scription and manipulation of information models. In-

formation Systems, 18(7):489–523.

Hoppenbrouwers, S. and Proper, H. (1999). Knowledge dis-

covery: De zoektocht naar verhulde en onthulde ken-

nis. DB/Magazine, 10(7):21–25. In Dutch.

ISO/IEC (1994). Information technology - open systems

interconnection - basic reference model: The ba-

sic model. International Standard ISO/IEC 7498-

1:1994(E), Information Technology & Management,

Geneva, Switzerland, EU. 2nd Edition.

Liang, T. (1994). The basis entity model: A fundamen-

tal theoretical model of information and information

processing. Information Processing & Management,

30(5):647–661.

Nonaka, I. and Takeuchi, H. (1995). The Knowledge Cre-

ating Company. Oxford University Press, New York,

NY, USA.

Proper, H. (1999). What is information discovery about?

Journal of the American Society for Information Sci-

ences, 50(9):737–750.

Roddick, J., Fule, P., and Graco, W. (2003). Ex-

ploratory medical knowledge discovery: experiences

and issues. ACM SIGKDD Explorations Newsletter,

5(1):94–99.

Weide, T. P. v. d. and Bommel, P. v. (2006). Measuring the

incremental information value of documents. Infor-

mation Sciences, 176(2):91–119.

Wiki (2006). Radiology. Wikipedia, the free encyclopedia.

http://en.wikipedia.org/wiki/Radiology.

WEBIST 2007 - International Conference on Web Information Systems and Technologies