TOOLKITS SUPPORTING OPEN INNOVATION IN
E-GOVERNMENT
Alexander Felfernig, Christian Russ
Institut fuer Wirtschaftsinformatik und Anwendungssysteme, Universitaet Klagenfurt
Universitaetsstrasse 65-67, A-9020 Klagenfurt
Manfred Wundara
Informationstechnologie Magistrat Villach
Rathausplatz 2, A-9500 Villach
Keywords:
e-government, open innovation, artificial intelligence, virtual advisors.
Abstract:
Today there exists a variety of efforts to bring public administration closer to its customers (citizens, en-
trepreneurs, etc.). This paper investigates the concept of open innovation, i.e. the integration of the customer
into the product creation process of a company, w.r.t. its applicability in the area of e-government. The concept
of open innovation is well known within the context of mass customizing products and services, i.e. produc-
tion and selling of customer-individual products and services under mass production pricing conditions. The
authors show how approaches from the area of artificial intelligence can be applied as tools for open innovation
in e-government.
1 INTRODUCTION
Henry Ford stated that his “customers can get ev-
ery car color they want as long as it is black”. This
is a perfect characterization of the mass production
paradigm. Ford was not confronted with any com-
petitors selling red or blue cars worldwide via on-
line shops. Customers did not ask for different equip-
ment variants, they were satisfied with one basic type
of car. However, times have changed and todays
enterprises are forced to continuously place product
innovations on the market. Within this context the
paradigm of mass customization (Anderson, 1997;
PineII et al., 1993) has been established propagating
the production and sales of customer-individual prod-
ucts and services under mass production pricing con-
ditions. However, from the customer point of view the
mass customization paradigm per se does not guar-
antee that the customer gets the products or services
he needs and he is looking for, it only guarantees
a highly variable product sortiment with competitive
development, production and distribution costs. Con-
sequently, the paradigm of mass customization has to
be enriched with the aspect of customer orientation,
i.e. the integration of the customer into the product
creation process of a company. The literature enti-
tles this approach to customer integration as open in-
novation (Tseng and Piller, 2003; Piller and Stotko,
2003), where companies actively provide customers
with means to articulate what they want and carefully
listen to customer requirements in order to generate
innovations that meet or maybe exceed the needs of
their customers. Tools provided to the customer in
this context are entitled as tools for open innovation in
the literature (VonHippel, 2001). Such tools support
user-centered dialogs by imposing personalized ques-
tions and provide explanations throughout the advi-
sory process. Results of the advisory process are pre-
sented to the customer taking into consideration his
knowledge level and interests. The result of an advi-
sory process is a set of identified products which fit
to the needs and wishes of the customer. In order to
guarantee the confidence in the result, a set of expla-
nations for the identified solution is given.
Mass customization of services in the public sec-
tor is majorly triggered by the increasing structural
complexity of society. On the one hand laws and reg-
ulations become more complex in order to consider
different social interest groups on the other hand cus-
tomers (i.e. citizens, entrepreneurs, etc.) are forced
to deal with a much larger and complex set of ser-
vices. While in the private sector tools for open in-
novation are increasingly provided for customers, in
the sector of public administration first steps are set
toward this direction. (Lenk et al., 2001) state that a
great potential exist to improve the interface between
administrative agencies and its customers
1
. Todays
1
Throughout the paper we use the term customer to de-
296
Felfernig A., Russ C. and Wundara M. (2004).
TOOLKITS SUPPORTING OPEN INNOVATION IN E-GOVERNMENT.
In Proceedings of the Sixth International Conference on Enterprise Information Systems, pages 296-302
DOI: 10.5220/0002648402960302
Copyright
c
SciTePress
e-government Web pages provide useful functional-
ities starting with simple presentations of organiza-
tional units and ending with different facets of on-
line transactions (e.g. renewal of a driving license
or a passport). Best practice examples provide life
event oriented views on the given services (Kavadias
and Tambouris, 2003). These approaches are useful
to increase the usability of large e-government por-
tals, however, none of those environments provides
services for effectively and interactively guiding and
assisting customers to find the services they require.
An entrepreneur searching for financial support for
research projects is forced to search through a large
number of different documents describing the pur-
poses of different research programs. Another en-
trepreneur trying to identify the best suited place of
business in a certain region is in the exactly same situ-
ation. In such situations virtual advisors can help cus-
tomers by providing personalized and explanatory di-
alogs with corresponding advisory results (e.g. well-
suited places of business for an entrepreneur includ-
ing an explanation why the recommended place was
chosen). This advisory support is provided via virtual
advisory channels (see Figure 1). A customer using
this channel contacts a virtual advisor to retrieve ser-
vices useful in his current situation (life event), i.e.
the customer initiates a dialog with the advisor sys-
tem.
information
services
advisor
services
organisational
knowledge and
services
citizen /
entrepreneur
public
administration
domain
knowledge
Figure 1: Open innovation in e-Government
Based on a set of stored user interactions stemming
from the dialog with the virtual advisor, the domain
expert (public administration in Figure 1) is now en-
abled to identify weaknesses in the actual advisory
process (e.g. which customer requirements signifi-
cantly correlate with the cancellation of the advisory
process?) and to identify customer requirements not
considered up to now (e.g. are there financial sup-
port programs frequently demanded by customers but
not provided in the actual portfolio?). An interpre-
tation of advisory sessions supports the domain ex-
note citizens, entrepreneurs, lawyers etc. having to interact
with public agencies.
pert when developing and maintaining advisor appli-
cations (this support is provided via customer feed-
back channel - see Figure 1). Feedback cycles are the
basic idea behind the concept of open innovation, i.e.
on the one hand to provide advisor functionality for
customers to alleviate the access to a complex prod-
uct or service, on the other to learn from customer
preferences in order to improve, change and extend
the product/service palette and the corresponding ad-
visory processes. Such automated feedback cycles are
extremely useful since politicians and other decision
makers in public administration can detect and re-
act much faster to open requirements existing in their
community, furthermore flexible improvements of ex-
isting advisory processes are enabled.
The remainder of this paper is organized as follows.
In Section 2 we sketch the different components of an
open innovation environment in the context of public
administration. Section 3 contains a description of
the used technologies, finally in Section 4 we discuss
related work. Section 5 concludes the paper.
2 SYSTEM ARCHITECTURE
Basically, an open innovation environment consists of
the components sketched in Figure 2 - the role of these
components and their interaction is discussed in the
following subsections.
Knowledge Acquisition. In order to be up-to-date,
the advisor knowledge base (product model, product
data, user profiles, interaction data) is continuously
adapted to the new situation (e.g. when new laws or
other regulations are introduced). In this context a
central modeling environment (Knowledge Acquisi-
tion component in Figure 2) is needed supporting the
effective maintenance of the advisor knowledge base.
When new laws are introduced the corresponding ad-
visor system should already be available to the cus-
tomer, i.e. the development- and maintenance time
for the advisor system is strictly limited.
As constraint representation languages of advisor
systems are hardly understandable for domain experts
and even for programmers themselves, development
environments for those systems must provide intu-
itive modeling concepts allowing the domain expert
to effectively develop and maintain advisor applica-
tions. In Section 3.1 the design of advisor knowl-
edge bases using the Unified Modeling Language and
its integrated constraint language OCL (Object con-
straint language) (Warmer and Kleppe, 1999; Rum-
baugh et al., 1998) is presented. UML is well known
and frequently applied in industrial software devel-
opment processes - the approach is intuitive and al-
leviates the integration of knowledge-based advisor
TOOLKITS SUPPORTING OPEN INNOVATION IN E-GOVERNMENT
297
techniques into industrial software development pro-
cesses.
citizen /
entrepreneur
public
administration
domain
knowledge
Knowledge
Acquisition
(UML, OCL,
Data Mining)
product
model
product
data
user
profiles
Virtual
Advisory
(Constraint
Satisfaction,
Personalization)
interaction
data
Figure 2: System architecture
Advisor knowledge bases consist of the following
parts:
• Product model: this model represents the basic
structure of the product including additional con-
straints (business rules) imposing restrictions on
the usage of different (sub-)products. An exam-
ple for such a product model is shown in the UML
(Rumbaugh et al., 1998) model of Figure 3, addi-
tional constraints (business rules) are shown in Fig-
ure 4 and Figure 5.
• Product data: the product model contains a set of
references to basic information sources which are
presented to the customer as part of the result of the
advisory process. If the result of a financial support
advisor is a certain program (e.g. the IST program
in Figure 3), additional information concerning this
program is stored in the underlying product catalog
(e.g. the attribute advisor contains a reference to a
specialized IST program advisor).
• User profiles: each registered user is described by
a set of dimensions describing his/her interests and
his/her knowledge level concerning different ser-
vice areas. This profile is used in order to person-
alize the advisory processes and the user interface.
• Interaction data: user interactions from advisory
sessions are stored here for the purpose of anal-
ysis. Following the idea of open innovation do-
main experts are enabled to analyze already given
customer interaction sequences. In order to sup-
port such feedback cycles the knowledge acquisi-
tion component includes data mining functionali-
ties supporting the derivation of association rules
and the application of different sorts of clustering
mechanisms.
Virtual Advisory. Virtual advisors support user-
centered dialogs by imposing questions and provid-
ing explanations - the result of an advisory session is
a set of proposed solutions with corresponding expla-
nations why this solution fits to the needs of the cus-
tomer. Within the context of virtual advisor applica-
tions customers have different knowledge levels and
interests concerning a set of provided services. For
example, customer A is very well informed about the
current situation w.r.t. to financial support of research
projects, while customer B knows nothing about this
domain and must get more detailed information about
the available programs. Consequently the whole di-
alog with the customer must be organized in a per-
sonalized fashion considering different interest- and
knowledge levels of customers w.r.t. the provided ser-
vice palette. Depending on the customers answers
and interaction behavior an intelligent user profile can
be developed. Which questions, hints and answers are
presented to the user, strongly depends on the infor-
mation stored in this profile.
3 CONCEPTS OF VIRTUAL
ADVISORS
3.1 Knowledge Representation
Our definition of an advisory task is based on the def-
inition of a configuration task (Felfernig et al., 2000).
Definition: Advisory task. An advisory task is
defined as a tuple consisting of the advisor knowl-
edge base ADV
KB
and a given set of customer re-
quirements ADV
REQ
, i.e. <ADV
KB
, ADV
REQ
>.
ADV
KB
consists of a description of product proper-
ties - concrete products (entries from a product cata-
log) are included in ADV
KB
as well. Additionally,
a set of constraints defines restrictions on allowed
combinations of instantiations of product properties
of ADV
KB
. The second part of an advisory task
is a given set of
customer requirements
ADV
REQ
which represent a set of instantiations of given cus-
tomer properties in the product model. ¤
Definition: Advisory solution. The goal of the
advising process is to identify a solution ADV
SOL
which fulfills the imposed customer requirements
ADV
REQ
where all the constraints of ADV
KB
are
fulfilled, i.e. ADV
REQ
∪ ADV
KB
is consistent. ¤
Service descriptions of the advisor knowledge base
are represented in UML (Rumbaugh et al., 1998) class
diagrams (a simple example for such a service de-
scription is shown in Figure 3 which will be used as
working example throughout the following sections).
The basic service structure is modeled using classes
(e.g. the class FinancialSupport), associations (e.g.
the association between the class Customer and the
ICEIS 2004 - SOFTWARE AGENTS AND INTERNET COMPUTING
298
EuropeanUnion
avgsum : 10000..1000000
%accepted : 2..8
topics : enum{’informatics’, ’economics’, ’agriculture’}
type : enum{’FundamentalResearch’, ’AppliedResearch’}
financing : enum{’loan’,’grant’,’both’}
<<Service>>
ERPFonds
avgsum : 10000..100000
%accepted : 20..40
topics : enum{’informatics’,’economics’}
type : enum{’AppliedResearch’}
financing : enum{’loan’,’grant’}
<<Service>>
FFF
avgsum : 300000..300000
%accepted : 30..30
topics : enum{’informatics’,’economics’}
type : enum{’AppliedResearch’}
financing : enum{’both’}
advisor : enum{’www.foerderkompass.at’}
<<Service>>
FWF
avgsum : 200000..200000
%accepted : 10..10
topics : enum{’Informatics’,’Economcis’}
type : enum{’FoundationalResearch’}
financing : enum{’grant’}
<<Service>>
IST
avgsum : 1000000..1000000
%accepted : 4..4
topics : enum{’informatics’}
type : enum{’FundamentalResearch’,’AppliedResearch’}
financing : enum{’grant’}
isthomepage : enum{’www.cordis.lu’}
advisor : enum{’www.cordis.lu\advisor’, ’www.fff.co.at\econtentadvisor’}
<<Service>>
Interreg
avgsum : 2000000..2000000
%accepted : 8..8
topics : enum{’economics’, ’agriculture’}
type : enum{’FundamentalResearch’,’AppliedResearch’}
financing : enum{’grant’}
<<Service>>
ProTec
avgsum : 100000..100000
%accepted : 30..30
topics : enum{’informatics’,’economics’}
type : enum{’AppliedResearch’}
financing : enum{’grant’}
<<Service>>
OENB
avgsum : 120000..120000
%accepted : 20..20
topics : enum{’economics’}
type : enum{’FundamentalResearch’}
financing : enum{’grant’}
<<Service>>
ERPOther
avgsum : 1000000..1000000
%accepted : 40..40
topics : enum{’informatics’,’economics’}
type : enum{’AppliedResearch’}
financing : enum{’loan’}
<<Service>>
Customer
companytype : enum{’gmbh’,’ag’}
risks : enum{’simpleextension’,’newtechnologies’,’highrisk’}
partners : Boolean
financing : enum{’loan’,’grant’, ’both’}
country : enum{’Austria’,’Germany’, ’Italy’}
topics : enum{’informatics’, ’econcomics’, ’agriculture’}
regionspecific : Boolean
<<RootComponent>>
FinancialSupport
avgsum : 0..10000000
%accepted : 1..100
topics : enum{’informatics’, ’econcomics’, ’agriculture’}
type : enum{’FundamentalResearch’,’AppliedResearch’,’Other’}
financing : enum{’loan’,’grant’,’both’}
<<Service>>
11 11
Figure 3: Service structure
class FinancialSupport which represents the relation-
ship between a set of imposed customer requirements
an the corresponding advisory solution ADV
SOL
),
and generalization hierarchies (e.g. FFF represents
a special kind of FinancialSupport).
Constraints defining restrictions on allowed combi-
nations of services are represented using the Object
Constraint Language (OCL) (Warmer and Kleppe,
1999) which is an integrative part of UML. Figure 4
shows two example constraints imposed on the ser-
vice structure shown in Figure 3. The first constraint
denotes the fact that financial support for customers
from foreign countries is restricted to cooperations
within the framework of European Union (financial
support of service type EuropeanUnion) projects. The
second constraint denotes the fact that simple exten-
sions to software components are not supported by the
current types of financial support programs
2
.
Basically, there are three different classes of con-
straints which are used very often when building
knowledge bases for virtual advisors.
• Requirement constraints: in some cases, the ex-
istence of a specific customer requirement (e.g.
country <> ’Austria’) requires the existence of a
specific service or a specific parametrization of a
service in the advisory solution (e.g. FinancialSup-
2
For reasons of readability the examples are kept simple
and do not regard the whole complexity of financial support
programs.
port.oclIsTypeOf (EuropeanUnion)).
• Compatibility constraints: certain combinations of
customer requirements (e.g. no project partners are
available and the customer stems from a foreign
country) can not occur in the same advisory session
- either these requirements are directly incompati-
ble or do not allow the identification of a solution
for the given advisory task.
• Resource constraints: parts of an advisory task can
be interpreted as a resource balancing task, where
some classes in the knowledge base produce some
resources and other classes act as consumers of re-
sources. E.g. when providing a more complex fi-
nancial support advisor, different potential partners
of a project consume resources in the form calcu-
lated costs - these costs must no exceed the overall
limit imposed by a given financial support program.
These three types of constraints represent frequently
used model restrictions in advisor knowledge bases.
For such types of constraints we provide a set of
graphical interfaces (constraint schemes) in order to
alleviate the constraint implementation for the devel-
oper of the advisor application. An example for a
requirement constraint represented on the graphical
level is shown in Figure 5. Note that constraints are
bound to a corresponding explanation - e.g. if the
customer has specified that (s)he is interested in lo-
cal funding opportunities and has an enterprise with
more than 500 employees, a corresponding hint could
TOOLKITS SUPPORTING OPEN INNOVATION IN E-GOVERNMENT
299
Figure 4: Example constraints
Figure 5: Graphical representation of requirement constraints
be presented indicating that no local funding is avail-
able for companies with more than 500 employees.
3.2 Personalization
As already mentioned, it strongly depends on the in-
formation stored in the user profile which questions,
hints and answers are presented to the customer dur-
ing an advisory session. Personalization concepts
play here an important role since one cannot assume
that the customer is well acquainted with the provided
services. On the basis of given interaction traces a
customer can be described using a set of abstract di-
mensions related to the provided products and ser-
vices (R. Schaefer, 2000).
Lets have a short look at the financial planning do-
main. Investment products can be categorized w.r.t.
to the dimensions profit, availability, and risk. De-
pending on the given customer requirements we can
determine a certain distribution describing the inter-
est of the customer w.r.t. the given dimensions. If
the customer articulates that no financial reserves are
available in the current situation, the dimension risk
will be of more importance, i.e. products with no fi-
nancial risks will be primarily presented to the cus-
tomer. A distribution of (profit=30, availability=60,
risk=10) describes a customer with nearly no readi-
ness to accept risks, with high interest in availabil-
ity of the investment and also an increased interest
in profitableness of the investment. The calculation
of such distributions is based on scoring mechanisms,
i.e. depending on customer answers a certain amount
ICEIS 2004 - SOFTWARE AGENTS AND INTERNET COMPUTING
300
of points is added to each dimension
3
.
Regarding our financial support scenario we in-
troduce the dimensions support (amount of financial
support which can be assumed if the project proposal
will be accepted) and acceptance (the probability of
the acceptance of the project proposal). For a con-
crete customer let us assume the following distribu-
tion {support: 30, acceptance: 60}, i.e. the cus-
tomer is much more interested in getting the project
proposal accepted than in getting a high amount
of financial support. Furthermore, let us assume
that the virtual advisor has determined four solutions
{S
1
, S
2
, S
3
, S
4
} for a given set of customer require-
ments (representing a request). For this set of solu-
tions and the given distribution {support: 30, accep-
tance: 60}, a personalized utility g(x) can be calcu-
lated for each solution alternative using the formula
g(x) =
P
n
i=1
e
i
(x)s
i
(x).
For each solution it is determined to what extend
this alternative is relevant for the different dimen-
sions. The utility g(x) is defined by s
i
(x) (rele-
vance of the alternative for the dimension i) and by
e
i
(x) which represents the customer interest in the
dimension i. The value g(x) for S
1
, i.e. g(S
1
)
= 30 * s
support
(S
1
) + 60 * s
acceptance
(S
1
). This
formula has to be applied for each of the given so-
lution alternatives - the result of the calculation is
a customer-specific ranking of the alternatives, e.g.
S
1
, S
3
, S
2
, S
4
. This ranking of advisory results is
one possible application of the presented personaliza-
tion concepts. The same concepts are used within the
scope of the advisory process in order to determine
the utility of different alternatives for questions, hints
or explanations. Furthermore, these concepts can be
used in order to determine default settings for parame-
ters which otherwise would have to be explicitly pro-
vided by the customer. This is helpful when a cus-
tomer does not know very much about the service do-
main or a parameter is of no interest for him.
3.3 Data Mining
In the context of open innovation, data mining tech-
niques (Berry and Linoff, 2000) can be applied in or-
der to determine patterns from previous interaction
sequences in the form of e.g. association rules or clus-
ters. The result of such an analysis can be used for
two different purposes. On the one hand the initial
version of an advisor application is not the last resort,
in many cases insufficient explanations and wrong re-
sults create an increasing frustration on the part of the
customer. In this context data mining techniques can
be applied for detecting areas in the advisor knowl-
edge base which are responsible for the bad system
3
A more detailed discussion on how to determine such a
distribution can be found e.g. in (R. Schaefer, 2000).
behavior. On the other hand customer requirements
can be analyzed with the goal to detect so called hid-
den requirements, i.e. requirements often articulated
by customers but not supported by the given service
palette. A simple example for such a result could
be that computer science projects proposed by insti-
tutions from foreign countries
4
having no domestic
partners are not supported by domestic financial sup-
port programs, i.e. topics=’informatics’ and coun-
try <> ’Austria’ and partners = false ==> ’no solu-
tion’. In the following the question has to be answered
whether this is intended or whether there exist oppor-
tunities to introduce such a financial support program
(maybe on an intra-regional level) in the future.
4 RELATED WORK
The
Webocrat
portal system proposed by (Paralic and
Sabol, 2001) supports customer integration by pro-
viding knowledge management (discussion forum),
content management (Web content management), and
polling functionalities (opinion polling room). The
citizen information helpdesk is a layer integrating
these functionalities by providing a text-based inter-
face supporting the search for support material. The
system supports personalization functionality by al-
lowing the user to declare his/her topics of interest.
Systems like Webocrat provide a first view on the de-
sign of future e-government systems.
(Vamos and Soos, 2003) discuss AI approaches in
the area of natural language understanding and case-
based reasoning bringing public administration closer
to its customers. This work can be seen as comple-
mentary to the concepts presented in this paper and
will be taken into consideration for future inclusion
into the presented toolsuite.
(Kavadias and Tambouris, 2003) present GovML,
a markup language to define structures for govern-
mental data and metadata. Compared to our work
GovML provides a domain-specific vocabulary for
the representation of public services and life events,
whereas our approach uses the concepts provided by
a domain-independent modeling language. The inter-
esting point here is how to integrate the concepts pre-
sented in GovML into a standard UML profile for the
area of public administration.
(Lenk et al., 2001) point out the great potential
which exists to improve the interface between ad-
ministrative agencies and the citizens. This interface
ranges from a mere human mediated form of interac-
tion to intelligent software agents. These agents im-
mediately provide information to users on where to
get online help and active support in specific cases.
4
In this example countries <> ’Austria’ are interpreted
as foreign countries.
TOOLKITS SUPPORTING OPEN INNOVATION IN E-GOVERNMENT
301
In this context our work is positioned on the level of
intelligent software agents providing intelligent infor-
mation and advisory services to customers.
In (Baar et al., 2000) experiences from the applica-
tion of OCL in industrial software development pro-
cesses are discussed. In principle, OCL seems to be
quite useful and software engineers and even domain
experts with a technical background are able to ap-
ply OCL for stating formal constraints on given ob-
ject models. Especially software engineers accepted
OCL because of the similarities of its syntax to object-
oriented programming languages. However, (Baar
et al., 2000) point out that more intuitive concepts
are needed in order to support effective modeling of
OCL constraints. In order to tackle this challenge
(Baar et al., 2000) introduce the notion of constraint
schemes which are parameterizable constraints which
can be differently instantiated depending on the ac-
tual situation. For example, a constraint schema could
restrict the number of objects of a class to an upper
bound - this upper bound is represented by a variable
which must be instantiated in order to instantiate the
corresponding OCL constraint. The graphical input
masks presented in this paper are very similar to the
constraint schemes presented by (Baar et al., 2000).
5 CONCLUSIONS
In this paper we have presented a toolsuite supporting
the concept of open innovation in e-government. Such
a toolsuite allows customers (citizens, entrepreneurs,
lawyers, etc.) to articulate their requirements, points
out and explains inconsistencies in given customer re-
quirements and provides needed service packages as
outcome of the advisory process. Using tracking data
from advisory processes, public agencies are enabled
to detect problem areas in advisor knowledge bases
and to generate innovations that meet the needs of
their customers. The construction of Virtual Advi-
sors requires a corresponding knowledge acquisition
workbench which allows the effective construction of
advisor knowledge bases. We have presented such
a workbench which is based on the Unified Model-
ing Language (UML) and the integrated Object Con-
straint Language (OCL).
REFERENCES
Anderson, D. (1997). Agile Product Development for Mass
Customization. McGraw-Hill.
Baar, T., Haehnle, R., Sattler, T., and Schmitt, T. (2000).
Entwurfsmustergesteuerte Erzeugung von OCL Con-
straints. In Mehrhorn, K. and Snelting, G., editors, In-
formatik 2000, 30. Jahrestagung der Gesellschaft fuer
Informatik, pages 389–404. Springer-Verlag.
Berry, M. and Linoff, G. (2000). Mastering Data Mining
Techniques. John Wiley & Sons.
Felfernig, A., Friedrich, G., Jannach, D., and Stumptner,
M. (2000). Consistency-Based Diagnosis of Configu-
ration Knowledge Bases. In Proceedings of the 14
th
European Conference on Artificial Intelligence (ECAI
2000), pages 146–150, Berlin, Germany.
Kavadias, G. and Tambouris, E. (2003). GovML: A
Markup Language for Describing Public Services and
Life Events. In Wimmer, M., editor, Knowledge
Management in Electronic Government: 4th IFIP
International Working Conference, KMGov 2003,
Rhodes,Greece, May 26-28, Lecture Notes in Com-
puter Science, volume 2645, pages 106 – 115.
Springer-Verlag.
Lenk, K., Traunmueller, R., and Wimmer, M. (2001). Elec-
tronic Business Invading the Public Sector: Consider-
ations on Change and Design. In 34th Hawaii Inter-
national Conference on System Sciences (HICSS-34),
volume 5. IEEE.
Paralic, J. and Sabol, T. (2001). Implementation of e-
government using knowledge-based system. In DEXA
Workshop, pages 364–368.
Piller, F. and Stotko, C. (2003). Mass Customization und
Kundenintegration. symposion, Duesseldorf.
PineII, B., Victor, B., and Boynton, A. (1993). Making
Mass Customization Work. Harvard Business Review,
Sep./Oct. 1993:109–119.
R. Schaefer, M. B. (2000). Ein allgemeiner Ansatz zur
Personalisierung von Webseiten mit Informationsdi-
ensten. In Adaptivitaet und Benutzermodellierung in
interaktiven Softwaresystemen.
Rumbaugh, J., Jacobson, I., and Booch, G. (1998). The Uni-
fied Modeling Language Reference Manual. Addison-
Wesley.
Tseng, M. and Piller, F. (2003). The Customer Centric En-
terprise. Springer, Duesseldorf.
Vamos, T. and Soos, I. (2003). Data Management and
AI in E-Government. In Wimmer, M., editor,
Knowledge Management in Electronic Government:
4th IFIP International Working Conference, KMGov
2003, Rhodes,Greece, May 26-28, Lecture Notes in
Computer Science, volume 2645, pages 230 – 238.
Springer-Verlag.
VonHippel, E. (2001). Perspective: User Toolkits for Inno-
vation. Journal of Product Innovation Management,
18:247–257.
Warmer, J. and Kleppe, A. (1999). The Object Constraint
Language - Precise Modeling with UML. Addison
Wesley Object Technology Series.
ICEIS 2004 - SOFTWARE AGENTS AND INTERNET COMPUTING
302
