74

Electronic Government, Vol. 3, No. 1, 2006

Change management in e-government: OntoGov

case study
Ljiljana Stojanovic*
FZIResearch Center for Information Technologies,
University of Karlsruhe,
Haid-und-Neu-Street, 10-14, Karlsruhe 76131, Germany
E-mail: [email protected]
*Corresponding author

Nenad Stojanovic
Institute AIFB,
University of Karlsruhe,
Karlsruhe 76128, Germany
E-mail: [email protected]

Dimitris Apostolou
PLANET S.A.,
Louise Riencourt 64,
Athens 11523, Greece
E-mail: [email protected]
Abstract: E-government systems are subject to a continual change. The
importance of better change management is nowadays, more important due to
the evolution of Europe towards a multicultural, more open and international
society with changing common values, increasing levels of education,
demographic involvement and adoption of new technologies. In this paper, we
show how semantic technologies may improve management of changes
regarding process knowledge in an e-government system. We consider change
management process as a continual improvement process. To improve the
usability of e-government services, we propose new methods for the semantic
service annotation as well as for semantic service discovery. Particularly, we
focus on new level of functionality such as verification of a service annotation
and refinement of search results.
Keywords: ontologies; e-government; change management.
Reference to this paper should be made as follows: Stojanovic, L.,
Stojanovic, N. and Apostolou, D. (2006) Change management in
e-government: OntoGov case study, Electronic Government, Vol. 3, No. 1,
pp.7492.
Biographical notes: Dr. Ljiljana Stojanovic is a Research Scientist in the
WIM Department of the FZI at the University of Karlsruhe. She received her
MSc degree in Computer Science from the Faculty of Electronic Engineering,
University of Nis, Serbia and Montenegro and a PhD degree from the
Copyright 2006 Inderscience Enterprises Ltd.

Change management in e-government

75

University of Karlsruhe for a thesis on Methods and Tools for Ontology

Evolution. She has participated in several R&D programs funded by the
European Commissions IST Programme related to e-learning, e-government
and Knowledge Management. Currently, she is a Technical and Scientific
Manager of the OntoGov project (IST). She has published more than 30 papers
in international journals and conferences in the areas of ontology engineering
and e-government.
Dr. Nenad Stojanovic is a Research Scientist in the Institute AIFB, University
of Karlsruhe. He received his MSc degree in Computer Science from the
Faculty of Electronic Engineering, University of Nis, Serbia and Montenegro
and a PhD degree from the University of Karlsruhe for a thesis on
Ontology-based Information Retrieval: Methods and Tools for Cooperative
Query Answering. He has participated in several R&D programs funded by
the European Commissions IST Programme related to ontologies, information
retrieval and Knowledge Management. He has published more than 50 papers
in international journals and conferences in the areas of ontology-based
information retrieval and Knowledge Management.
Dr. Dimitris Apostolou is a Manager in PLANET S.A., a management
consultancy operating in South-East Europe. He is a Technology Consultant in
the areas of e-government, e-business, e-learning, decision support systems
and Knowledge Management. He has participated in more than nine research
and development projects. He has coordinated the INKASS EU-funded
project and he is currently a Coordinator of the OntoGov project. He has
published more than 35 papers in international journals and conferences.
He received his MSc degree in Chemical Engineering from NJIT (USA), MS
degree in Information Technology from UCL (UK) and the PhD degree in
Electrical and Computer Engineering from NTUA (Greece).

Introduction

The Semantic Web (SW) has been in the focus of the AI community for the last five
years. However, after years of intensive research and impressive scientific results, what
the SW now really needs is real world use cases, to demonstrate its added (business)
value. Moreover, the full application potential of some SW technologies, such as SW
Services and Rules has been neglected due to a lack of large-scale testing domains.
Finally, the next application-driven research challenges for the SW can be defined only
through the feedback from real use cases. Therefore, the SW requires a large, dynamic,
heterogeneous and shared information space to be effectively evaluated.
On the other hand, the domain of e-government is unique because of its enormous
challenge to achieve inter-operability, given the manifold semantic differences of
interpretation of, for example, law, regulations, citizen services, administrative processes,
best-practices and, last, but not least, the different languages to be taken into account
within and across regions, nations and continents. Setting-up seamless e-government
services requires information integration as well as process integration involving a
variety of objects with specific semantics (Klischewski, 2004).
Therefore, the combination of these two domains seems to be quite natural: the
e-government domain can provide an ideal test bed for existing SW research, and SW
technologies can be an ideal platform to achieve the vision of a knowledge-based,

76

L. Stojanovic, N. Stojanovic and D. Apostolou

user-centric, distributed and networked e-government (EU Report, 2004a). Moreover,

due to its open architecture, e-government provides a palette of new research questions
for SW, such as interportal search (e.g. searching for additional resources on other portals
to reply to a primary user or agent request). E-government further exhibits some
remarkable characteristics which make it more demanding on one hand, but also more
promising, on the other hand, than common e-business scenarios; to mention but a few
such characteristics: high degree of formality of key areas (law); extreme requirements to
come to the same decisions in similar situations; high demands with respect to security,
privacy and trust; sometimes extremely long-running process instances (e.g. in urban and
regional planning); sometimes extreme information imbalances between stakeholders, as
well as many different stakeholders in the same process (e.g. citizen versus city council,
county council and federal government) and many more.
In this paper, we discuss how SW technologies may improve change management in
e-government. Indeed, the increasing complexity of e-government services demands a
correspondingly larger effort for management. Today, many system management tasks,
such as service reconfiguration due to changes in the law, are often performed manually.
This can be time-consuming and error-prone. Moreover, it requires a growing number of
highly skilled personnel, making e-government systems costly.
Owing to its formal nature, semantic technologies seem to promise support for
resolving drawbacks in existing change management systems for e-government, as
stressed in the recently performed analysis of the research challenges for e-government in
the next decade (EU Report, 2004b). By using ontologies for modelling e-government
services not only the consistency preservation and change propagation can be achieved,
but also more important a qualitatively new level of functionalities can be provided. The
approach has been implemented within the OntoGov1 project, whose main objective is to
develop a semantically enriched platform that will facilitate the consistent configuration
and reconfiguration of e-government services.
This paper is organised as follows. The requirements for change management in
e-government are discussed in Section 2. The set of ontologies needed for a better
management of e-government services is introduced in Section 3. Advantages of using
ontology-based change management are discussed in Section 4. Before we conclude, we
present an overview of the related work.

Motivating example

To make the description of the approach more understandable, we define here the basic
structure of an e-government system and give a motivating example that will be used
throughout the paper. There are four basic roles played by the actors in an e-government
system:
1

politicians who define a law

domain experts who define processes for realising a law

programmers who implement these processes

end-users (applicants) who use e-government services.

The dependencies between them are shown in Figure 1.

Change management in e-government

Figure 1

77

As-is situation

Domain experts have a key role. They possess a very good knowledge about the
e-government domain. This knowledge is needed for the design of a public service.
It includes the legislation that a service is based on, the respective law, related decrees,
directives, prerequisites, etc. On the basis of the interpretation of a law, a domain expert
describes a service as a sequence of activities that have to be done, which represents a
business process. For example, the generic schema for the public service for issuing
(renewal) a driving licence is realised through the following five activities:
1

application

verification/qualification

credential issuance

record management and

revenue collection.

In the application activity, all the necessary application data/documents are provided by
an applicant. In the next activity, the provided information/documents are verified
(e.g. validity and liquidity of a credit card) and are qualified by testing whether the
applicant meets the qualification requirements. In the issuance activity either a permanent
or a temporary credential document (i.e. driving licence) is issued. The record
management activity ensures the ongoing integrity of the driving licensing and control
record. Finally, the required fee is charged from the applicants bank account. Each
activity requires some inputs and produces some outputs. It can be executed only when
its pre-conditions are fulfilled and it has post-conditions that define the next activity in a
conditional manner. In the case of the application activity of the driving licence service,
inputs include a birthday certificate, the output is an application form, the pre-condition
is that the applicant is older than 16 and the post-condition is that all fields in the
application form are filled. Further, each activity can also be decomposed into several
subactivities or can be specialised.
The crucial activity is the verification/qualification as it reflects the constraints
contained in the law. For example, it implements a rule that a person younger than 16
cannot apply for issuing the driving licence, whereas for motor cars (category B) the
minimal age is 18. From the business process management point of view, the law can be
treated as the business rule required to achieve goals of an organisation (defined by its
business policy).

78

L. Stojanovic, N. Stojanovic and D. Apostolou

Owing to the changes in the political goals of a government, changes in the

environment and changes in the needs of the people or due to the possibility to organise
regulations in a better way, the politicians might:
1

make the revision of a law by accepting an amendment

enact a new law and

even repeal a law.

In the case of a new amendment, the domain expert must understand the changes in
the law caused by the amendment; locate activities/services in which this law has to be
implemented, and translate changes into the corresponding reconfiguration of
the business process. Let us continue the example with the driving licence. Recently, the
German law that regulates issuing driving licences has been changed, so that foreigners
from non-EU countries must have the German driving licence, although they have a
domestic licence. Let us analyse which changes in the existing business process for
issuing the driving licence will be caused by this change in the law. For each change, we
discuss the role that an efficient change management system should play. First of all, the
domain expert should locate a business process and the corresponding activities that
should be modified due to this change in a law. This is a time-consuming action if it is
performed in a non-systematic way. Therefore, an efficient change management
approach should inform the domain expert on these activities automatically. It means that
each business activity must contain a reference to a chapter/paragraph/article/amendment
of a law it implements. For example, the activity verification/qualification of the driving
licence service is based on Chapter 2, Paragraph Mindestalter in the Law Bundesgesetz
ueber den Fuehrerschein.
After finding the service that has to be modified, the domain expert has to decide how
to do that. She can specialise this service in the new one or she can adapt it to include
new requirements. Let us assume that the domain expert made a decision to generate a
specific driving licence service for foreigners. This service should not be generated from
scratch. Rather, it should be a specialisation of an already existing driving licence
service. The domain expert has to change the pre-conditions of this new service, since it
is only for foreigners from non-EU countries. This automatically causes a change in the
pre-conditions of the original service,6 as the pre-conditions of two different services that
provide the same functionality must be disjoint. Only in this way the run-time system
will know which service to execute. It is clear that when the pre-conditions are
semantically defined, the judgement about the inclusion relation among them can be
done automatically.
Further, the verification/qualification activity of the new service requires checking
whether a foreigner already has a domestic licence. Therefore, a new input for that
activity is necessary. Since each input has to be supplied, this change is propagated to
the previous activity, that is, the application activity, which is responsible for the
interaction with an applicant. It means that the activity has to deliver (as its output)
the information about the domestic licence, the validity of which should be tested in the
verification activity. Consequently, the application activity of the new service needs an
additional input compared to the original service.
Obviously, different changes in a law have different consequences in the existing
services. We briefly discuss one more example. Recently, the German law that regulates
issuing driving licences has been changed, so that teenagers older than 17 can obtain a

Change management in e-government

79

(temporary) licence for motor cars if they pass the exams and if they drive with a person
who is older than 25, has the driving licence for more than five years, and has scored less
than 20 negative points7 in the last five years. In that case, the older person must have a
licence for co-driving. This change in the law requires changes in the post-conditions of
the verification/qualification activity: instead of approval and non-approval of the
licence, it can be temporarily approved. Further, the credential issuance activity has to
generate an additional output, as the new co-driving licence should be printable as well.
An efficient change management system should enable the domain expert to perform all
these changes efficiently (e.g. to make a minimal set of additional changes) and to ensure
the overall consistency of the reconfigured service automatically (e.g. to prohibit that an
activity has two contradictory preconditions).
An ontology-based change management system that fulfils the above mentioned
requirements has been developed within the OntoGov project. The OntoGov system
ensures harmonisation of requests for changes, resolution of changes in a systematic way
and their consistent and unified propagation to the all-dependent artefacts. More
information can be found by Stojanovic et al. (2004). In the rest of this paper, we focus
on additional advantages that can be achieved by using ontologies.

OntoGov ontologies

For e-government initiatives to succeed, in addition to modernising the front office,

attention should be paid to streamline, reorganise and support the back-office processes
of public administrations that provide e-government services to citizens. Furthermore,
actions should be taken to limit the loss of critical knowledge assets during the life cycle
of e-government services. The overall objective of the OntoGov project is to develop,
test and validate a semantically enriched platform that will facilitate the consistent
reconfiguration of e-government services.
Therefore, the first task is to define ontologies for modelling e-government services.
On the basis of the analysis of the existing standard for Semantic Web Services
(i.e. OWL-S and WSMO) and by taking into account the e-government specific
requirements, we have defined a set of ontologies. They can be clustered in the following
way (see Figure 2):

Meta Ontologies

Domain-Oriented Ontologies and

Administration Ontologies.

Figure 2

Clusters of e-government ontologies

80

L. Stojanovic, N. Stojanovic and D. Apostolou

The Meta Ontologies define the schema, that is, the language for modelling
the e-government services. The Domain-Oriented Ontologies model the concrete
e-government services and all data relevant for these services. Since the goal of
the OntoGov project is to enable better management of e-government services, we have
introduced the Administration Ontologies.
The Meta Ontology Cluster contains general ontologies that do not change from one
deployment to another. It consists of the following ontologies:

the Legal Ontology defines the structure of the legal documents, which includes
paragraphs, sections, amendments, etc.

the Organisational Ontology models an organisation by defining its organisational

units, roles, persons, resources, etc.

the Lifecycle Ontology describes the information flow and the decision-making
process in the public administration. Each design decision refers to the entities either
from the Legal Ontology or from the Organisational Ontology or to other design
decisions, since they drive the decision

the Domain Ontology contains domain-specific knowledge

the Process Ontology describes the elements for modelling the process flow.
It includes the Domain Ontology for defining inputs and outputs as well as the
Lifecycle Ontology for explaining reasons that motivate the decisions

the LifeEvent Ontology models the categorisation of the e-government services

the Profile Ontology contains metadata about e-government services and includes all
previously mentioned ontologies.

The dependency between these ontologies is shown in Figure 3. The Profile Ontology
and the Process Ontology are defined based on the corresponding OWL-S ontologies by
taking into account the e-government specificities such as a reference to the law that is
modelled through the Legal Ontology.2 The Domain Ontology defines terminology
used in the e-government domain (e.g. type of documents such as passport). The
Organisation Ontology is defined to take into account experiences from the business
process modelling and reengineering, since the change in the organisational structure can
cause the changes in the process model. The LifeEvent Ontology is specific for the egovernment domain and it is defined to support better searching for e-government
services. The Lifecycle Ontology is defined to help the domain expert to introduce the
changes in the service description and to document the reasons for these changes.
The cluster of Domain-Oriented Ontologies contains a set of ontologies that are
structured in accordance with the specific domain, for example, pilot. These ontologies
are specialisation3 of ontologies belonging to the cluster of Meta Ontologies or other
domain-oriented ontologies.
For example, at the government level we may define the Legal-Federal Ontology
based on the Legal Ontology that belongs to the cluster of Meta Ontologies
(see Figure 3). It contains the entities representing the laws that hold at federal level.
Each federal state has its own laws. Therefore, the Legal-State Ontology may be a
specialisation of the Legal-Federal Ontology (since a state must satisfy all federal laws)
by extending it with the knowledge related to the federal state laws. Further, each
municipality may create its Legal-Municipality Ontology that extends the Legal-State

Change management in e-government

81

Ontology with some regulations. This example is shown in Figure 4. We note that there
are no constraints regarding the depth of the specialisation. However, our approach is
currently limited to include entire models rather than including subsets. Also, when a
model is reused, information can only be added and not retracted.
Figure 3

The cluster of Meta Ontologies

Figure 4

Specialisation of the ontologies describing legal aspects

The main ontology in the cluster of Domain-Oriented Ontologies is the so-called Service
Ontology. Each e-government service is represented by one Service Ontology. A Service
Ontology is an instantiation of the Profile Ontology and it contains specialisations of all
ontologies included in the Profile Ontology. The profile part for the concrete service
Announcement of move is shown in Figure 7. The graphical representation of a process
model (that is also serialised as an ontology) is shown in Figure 8.
We note that the specialisations of the Legal, Organisational, Domain and
LifeEvent Ontologies may be shared between several Service Ontologies. It means that
for each particular governmental institution the domain experts have to define their own
specialisation of the Legal, Organisational, Domain and LifeEvent Ontologies by taking
into account the specificities of this public administration. Moreover, they have to reuse

82

L. Stojanovic, N. Stojanovic and D. Apostolou

as much as possible of already existing knowledge. For example, to define their own
ontology for legal aspects they should reuse the ontology representing the law at the state
level (or at least at the federal level) and not to start directly from the Legal Ontology.
This will speed up the ontology development process and will increase the
inter-operability.
On the other hand, the specialisations of the Lifecycle, Process and Profile Ontologies
are specific for each concrete service. This specialisation is done by creating instances
and property instances of corresponding concepts defined in some of the included
ontologies. It means that for each e-government service we have exactly one
specialisation of these three ontologies, as each e-government service has its own profile,
process model as well as life cycle. We note that a Service Ontology may include other
Service Ontologies.
Finally, the cluster of the Administration Ontologies contains two types of ontologies:

ontologies used for tracking changes in the description of the e-government

services

ontologies used for deployment and execution of the e-government services.

For tracking changes we have defined the Evolution Ontology. The Evolution Ontology is
a model of changes that can be applied to the service description. This formal model
enables better management of the changes, as only a common understanding of the
changes enables the synchronisation between the evolving ontology and the dependent
artefacts that have to incorporate or adapt to those changes.
Moreover, for each particular service there exists exactly one Evolution Log Ontology
that tracks the history of changes applied to the description of the considered service. It
enables the recovery from failure since it makes possible to undo and redo applied
changes as needed. This ontology includes the Evolution Ontology,4 since the Evolution
Ontology defines the structure of the log file. However, it only refers
the Service Ontology, which means that all relationships to the entities from the Service
Ontology are modelled as attributes whereas values are the unique identifier of the entity
from the Service Ontology that is changed. The main reason is that it may be possible that
referenced entities do not exist anymore (e.g. an atomic service is removed).
For service deployment and execution the WSOR Ontology is introduced. It defines
information needed for the service deployment. The instantiation of this ontology is
called the WSOR Service Ontology and it has to be specified for each particular service.
It models dependencies between run-time data and a concrete web service that has to be
invoked. Moreover, it establishes mapping between the service description and the
WSDL description of a web service. Similar arguments as previously can be applied
here for the dependencies between the WSOR Ontology, the WSOR Service Ontology and
the Service Ontology.

Advantages of using ontologies

In this section, we discuss the advantages of ontology-based change management.

We begin this section by describing the benefits from a modelling point of view
(see Section 4.1). Then, we elaborate the benefits from the usage point of view (see
Section 4.2). More information about other aspects of ontology-based change
management can be found by Stojanovic et al. (2004).

Change management in e-government

83

4.1 Verification
The basic requirement for a management system is that it has to be simple, correct and
usable for domain experts. Note that they are responsible for keeping semantic
description of services up-to-date and do not need to be experienced ontology engineers.
Thus, a management system must provide capabilities for the automatic identification of
problems in the (description of the) semantic description of e-government services and
ranking them according to the importance. When such problems arise, a management
system must assist the domain experts in identifying the sources of the problem, in
analysing and defining solutions for resolving them. Finally, the system should help in
determining the ways for applying the proposed solutions.
In this section, we define the procedure for finding the weak places in the
description of the e-government services by considering the semantics on underlying
ontology model that is introduced in the previous section. The procedure is focused on
discovering inconsistencies in a service description. When we designed this support,
we assumed that the update would be only a partially automated process rather than a
fully automated process. For example, we do not want to update service description
automatically, but rather to notify the domain experts about problems. It is up to them to
decide how to resolve those problems. Our experience shows that this assumption is
reasonable. In the e-government domain, certain tasks could be automated, while other
tasks could be supported, but not fully automated.
The approach requires the explicit specification of consistency. We define an
ontology-based e-government service description as a consistent one if and only if it
satisfies the following conditions:

it is ontology consistent5 and

it satisfies a set of consistency constraints defined for the underlying ontology

model.

The set of constrains heavily depends on used ontology language. The consistency of
OWL ontology language is defined by Stojanovic (2005).
The consistency constraints of e-government service descriptions (see Section 3) are
defined to take into account specificities of the ontologies used for modelling, as they
represent the language for describing services. This type of constraints is called the
user-defined consistency constraints. They are users requirements that need to be
expressed outside the ontology language itself. Two types of the user-defined
consistency conditions are identified:

General conditions that are applicable across domains and represent best design
practice or modelling quality criteria such as redundancy, misplaced properties,
missing properties, etc.

Domain-dependent conditions that take into account the semantics of a particular

formalism of the domain. The full set of constraints for e-government services is
given by Stojanovic et al. (2004). For example, one constraint specifies that all
entities in the process model must be connected. A set of rules used for verifying
this constraint are shown in Figure 5.

84

L. Stojanovic, N. Stojanovic and D. Apostolou

Figure 5

Verification based on the domain-specific constraints. A part of consistency rules is

depicted in the left part. The right part shows the process model that does not
satisfy the rules (see black activity)

We note that the general conditions are mainly applied on a service profile, whereas the
domain-dependent conditions treat the process model. In the rest of this section,
we discuss how general conditions can be used for the verification of an ontology-based
description of e-government service. Owing to lack of space we omit here the discussion
of domain-dependent conditions.

4.1.1 General conditions

The quality of service description is of vital importance for a more accurate and timely
decision making. Indeed, the service description may be in consistent, but it may contain
some redundant entities or can be better structured with respect to the e-government
domain. Three quality criteria are defined in the following way:

Compactness: a service description is not compact or it is redundant if it contains

more metadata than it is needed and desired to express the same idea. To achieve
compactness (and thus to avoid redundancy), the description has to comprise the
minimal set of the metadata without exceeding what is necessary or useful. The
repetition of the metadata or the usage of several metadata with the same meaning
only complicates maintenance and decreases the system performance.

Completeness: a service description is incomplete if it is possible to extend the

description only by analysing the existing metadata in the description to clarify its
semantics. It means that the description is not finished yet and requires that some
additional metadata has to be filled in.

Aggregation: a service description is aggregative if it contains a set of metadata that

can be replaced with semantically related metadata to achieve a shortened
description, but without producing any retrieval other than the original description.

To clarify the meaning of the quality criteria here we give a short example. It is shown
in Figure 6. For example, the concept hierarchy and the property hierarchy from the
domain ontology can be used to refine description. Figure 6 represents the incompact
description because the service is annotated, after all, with the concept Person and its
subconcept Female. When someone searches for all services about Person, she
searches for the services about all its subconcepts (including Female) as well.
Consequently, she gets this service (minimum) twice. Moreover, such a description
introduces an ambiguity in the understanding of the content of a service, which implies
problems in knowledge sharing. Let us examine the meaning of the description of a

Change management in e-government

85

service using the set of metadata Person, Female and Passport. Does it mean that
the knowledge item is about issuing passport only in females, or in all persons? When the
second answer is the right one, then this service is also relevant for the treatment of male
persons. This implies new questions: is the description using metadata Female an error,
or the metadata Male is missing? Anyway, there is an ambiguity in description, which
can be detected and resolved by using the proposed approach.
Figure 6

Refinement based on the general conditions. The domain ontology is depicted in the
left part. The right part shows downward the initial metadata, and the improved
semantic description

To prevent this, a service should be annotated using as special metadata as possible

(i.e. more specialised subconcepts). In this way, the mentioned ambiguities are avoided.
Moreover, the maintenance of the descriptions is also alleviated because the description
is more concise and because only the changes linked to the concept Female can
provoke changes in the description.
The second criterion is related to the completeness of metadata. It is computed based
on the structure of the domain ontology. For example, one criterion is the existence of a
dependency in the domain ontology between the domain entities, which are already used
in the annotation. The second example in Figure 6 contains concepts with many
relationships between them (e.g. properties influences and determines exist between
concepts Visa and Marriage). The interpretation is ambiguous as it is a question
whether the services are:
1

how marriage influences issuing of a visa or

how type of visa determines services needed for a marriage.

86

L. Stojanovic, N. Stojanovic and D. Apostolou

To constrain the set of possible interpretations, the annotation has to be extended with
one of these properties.
This problem is especially important when the repository contains a lot of services
annotated with the same concepts because the search retrieves irrelevant services that use
certain concepts in a different context. Consequently, the precision of the system is
decreased.
The third pattern for the refinement occurs when a service is described with all
subconcepts of one concept (e.g. concepts Female and Male as shown in the third
example of Figure 6). From the searching for services point of view, it is the same
whether a service is annotated using the combination of the concepts (e.g. Female and
Male) or using only the parent concept (e.g. Person). It is obvious that the second
case of annotation makes the management much easier. Moreover, as the standard
approaches to the ranking results of querying exploit conceptual hierarchies, for example,
in a querying for persons a service annotated using Female and Male will be placed at
the same level as a service annotated using only one of these concepts. However, it has to
be ranked on the top level (level of the concept Person) because it covers all subtypes
of the concept Person.

4.2 Search
Owing to increase in the automation of e-government processes, the number of available
public services proliferates. To speed up the development process and to enable the
inter-operability between different public administrations, the e-government services
should be reused as much as possible. One of the crucial problems is how to find the
most suitable service for a given user need.
Service discovery is the process of locating services that can be used to request a
service that fulfils some user needs. Currently, existing service discovery techniques use
simple interface- or attribute-based matching. Service discovery is effectively done at a
syntactic level. It is well known that syntactic level matching and discovery is inefficient.
From the user point of view, there is the problem which terms or keywords to use when
searching for services. Simple keyword queries are valuable in situations where users
have a clear idea of what they are searching for, and the information is well defined. It
does not work in e-government, where the viewpoints and the knowledge
levels of the service provider and service consumer may be completely different.
Therefore, some mechanism for establishing a shared understanding is needed. Second,
simple keyword searches cannot deal with synonyms (passport and pass),
abbreviations (world wide web and www), different languages (passport (English)
and der Reisepass (German)) and even morphological variations (e-government and
eGovernment), not to mention the context.
This problem is especially important in the e-government domain, as the expert
knowledge about what regulatory rules apply for obtaining a service and how to obtain
the requisite information is now expected from an average citizen. Moreover, often
government services span agency boundaries that require services from not just one
agency but from several. For example, new business registration services may require
services from the Department of Commerce, Department of Revenue, Division of
Commercial Recording, Division of Community Affairs, Department of Environmental
Protection, Department of Public Health and Safety, etc. To provide a cohesive, seamless

Change management in e-government

87

and comprehensive inter-agency and inter-governmental service to the citizen, syntactic

service discovery is not enough. Users are expected to find services without first
acquiring the knowledge about them.
Thus, there is a need to discover services in a semantic manner. The service
discovery problem is now defined as a problem of matching formally described user
requests with service functionality satisfying these requests. Reasoning is used to infer
information about the capabilities and functionality of different services. Even though
semantic descriptions of services can be exploited to improve the process of locating
services that achieve a given requester goal, it is still hard to locate the relevant services.
The main problem is that the users are not experienced ontology engineers and they have
problems to specify ontology-based queries. Numerous existing proposals for discovery
follow the approach where service descriptions are expressed by concept expressions in
description logics, and description matching is performed by well-known description
logic inferences. However, these approaches do not always produce results one might
intuitively expect, due to a gap between the formal semantics of service descriptions and
human intuition (Grimm et al., 2004).
While a number of proposals for automating the discovery of services are available
(Grimm et al., 2004; Li and Horrocks, 2003; Paolucci et al., 2002), these hardly consider
the fact that a users query is just an approximation of his information need. Thus, the
main issue in realising an efficient service discovery system is to treat the discovery as an
exploratory process, in which a user should develop his need incrementally. Indeed,
from the more general problem, search for information in an information repository is
known that:

a user is usually unfamiliar with the content of the information repositories.

To avoid making an overspecified long query that retrieves zero results
(i.e. a failing query), the user starts search with a short query and tries to exploit
the repository in several subsequent refinement steps

a user often has ill-defined information need. He starts search by assuming what
can be the right information, but often, by exploring the resulting list, he redefines
what he is actually searching for.

Therefore, a user should develop his need for a service incrementally, in the so-called
step-by-step manner, to efficiently find the most suitable service. Such a discovery
process we will call step-by-step service discovery (Stojanovic, 2005).
In the nutshell of the step-by-step searching is the process of expanding or redefining
the initial query to obtain more relevant results the so-called query refinement process.
However, existing methods for query refinement seem to be inadequate for real world
usage (Stojanovic, 2004) and especially for service discovery, as they usually return a
long list of mainly irrelevant refinements. Indeed, the recent experimental studies in the
interactive query refinement have shown that only one-third of the terms derived from
document relevance feedback were identified by users as useful for refining their
searches (Campbell, 2000). In other words, users are overloaded with refinement
information, similarly to overloading with search results6 in an information retrieval task.
In our previous work, we have developed a comprehensive approach for the
refinement of keyword-based queries, the so-called conceptual query refinement
(Efthimiadis, 2000). To deal with the meaning of a keyword-based query, we have
developed an ontology that defines the conceptual space in which a query can be

88

L. Stojanovic, N. Stojanovic and D. Apostolou

interpreted, so that the query refinement process is performed on the level of the query
model (i.e. meaning of a query). Indeed, a term can be considered as a refinement of a
query, if and only if it can be inferred from the query model. The ontology is instantiated
from the results of a query, which ensures the relevance of generated refinements for a
users information need. The logic-based nature of the refinement enables a variety of
additional services that enrich the query refinement process, like cooperative answering
(Stojanovic and Stojanovic, 2004). In that way the approach goes beyond refinement of a
users query towards the refinement of the users information need.
In the rest of this section, we present how this approach can be applied for the more
efficient, discovery of e-government services. We focus on the service profile, as it
provides a representation of properties and capabilities that can be used by service
requesters to specify their needs and service providers to advertise their services for
what they do. Figure 7 shows an example of e-government service profile for the
service Announcement of move. The process model is shown in Figure 8. This service
is classified as of high potential for European e-government improvement (Klischewski,
2004), as is typically involving various public and private institutions.
Figure 7

The instantiation of the profile ontology for the Announcement of move service

Change management in e-government

Figure 8

89

A process model of the Announcement of move service

Search for e-government services (see Figure 9) is realised by combining three different
types of conditions:

Query-by-example (QBE) it is used for specifying conditions on the service

definitions (e.g. 1 in Figure 7), supplying the constraints on various fields. For
example, the user may search for services that are ready to be deployed (cf. the field
Status is shown in Figure 9).

Reasoning using class hierarchy the user can specify the type (i.e. classification7
cf. 2 in Figure 7) of services. Subsumption reasoning is used to locate services that
are more specific than specified. For example, even though the user issued a search
for services about residential affair (cf. Classification in Figure 9), a result to this
search included services classified in the field of 10_IdentificationCertification
and/or #20_Moving due to hierarchical relations (i.e. the class
17_ResidentialAffairs is a superclass of them).

Conceptual Query Refinement the user defines keywords specifying the relevant
terms that the service description must contain. The refinement system takes as the
input the results of keyword-based search; whereas results are service descriptions
(see Figure 7). The system calculates refinements based on the values for the
property hasDescription (cf. 3 in Figure 7). It generates a set of possible extensions
of the original query using the method described in Section 3. For example, if a user
puts a keyword-based query deregistration, the refinement system produces a set
of refinements that indicate a set of activities that are related to the concept
deregistration, such as address (as indicated in Figure 7), vehicle, weapons,
etc. Note that these refinements are generated from the textual descriptions of
services and are semantically related to the original query, that is, the refinement
system does not take into account only the co-occurrence between words, but
more important their semantic relatedness. More information about semantic
relatedness can be found by Efthimiadis (2000).

90

L. Stojanovic, N. Stojanovic and D. Apostolou

Figure 9

Screenshot from the OntoGov registry (http://wim.fzi.de:8080/ontogov/ontogov.jnlp)

(1) depicts the QBE, (2) depicts search using a class hierarchy and (3) denotes
conceptual query refinement

Related work

In Klischewski (2004), semantic problems in e-government are identified. They are

prerequisite for discussing requirements for the application of SW technologies. Even
though a research agenda to guide and to support the application of SW technologies in
e-government is available, until now there are still very few approaches that tackle
theoretical, technical and application aspects of the usage of SW methods for
e-government problems. This paper is one step towards this direction.
OWL-S uses an upper ontology to semantically describe web services. Even though
OWL-S assumes manual annotation of web services, we believe that annotation in the
real world will be a non-trivial task, without some degree of automation. The Meteor-S
system (Patil et al., 2004) primarily aims on providing a semi-automatic approach to
matching elements in WSDL to ontologies. Hess and Kushmerick (2003) talks about
using semantic metadata to semi-automatically categorise web services into predefined
categories making the service discovery simpler. It uses machine learning techniques for
categorisation. Our work presents an approach for not only adding semantics to
e-government services, but also more important to verify this annotation. As known
to the authors, none of the existing systems for web services offer support for
(semi-)automatic improvement of annotation. However, this support is very important as
one method for reducing the cost is to automate aspects of the evolutionary cycle when
possible.
A number of proposals for using description logic for automatic discovery of services
are available (e.g. Grimm et al., 2004; Li and Horrocks, 2003; Paolucci et al., 2002).
However, our approach differs from the above presented ones by introducing the

Change management in e-government

91

conceptual level, which is used for the analysis of a users query. It enables more
efficient refinement process. Moreover, our inference-like search is a unique feature, as
well as the cooperativeness enabled by it. The important difference is the process-driven
nature of our approach, especially the existence of the refinement ranking phase.

Conclusion

E-government systems are subject to a continual change. The importance of better

change management is nowadays more important due to the evolution of Europe towards
a multicultural, more open and international society with changing common values,
increasing levels of education, demographic involvement and adoption of new
technologies. It is especially true for the new EU countries, since the European
integration has paved the way for new legislation, regulations and corresponding changes
that affect the way public administrations in the enlarge Europe are organised and
operate.
It is clear that ad hoc management of changes in e-government might work only for
particular cases. To avoid drawbacks in the long-run, the change management must be
treated in a more systematic way. In this paper, we presented an approach for
ontology-based change management. Our approach goes beyond a standard change
management process; rather it is a continual improvement process. To improve the
usability of e-government services with respect to the needs of users, we proposed
methods for the semantic service annotation as well as for semantic service discovery.
The formal models of e-government services offer new level of functionality such as
verification of annotations and refinement of search results.

Acknowledgement
The research presented in this paper was partially financed by EU in the project IST
PROJECT 507237 OntoGov and by BMBF in the project SemiPort (08C5939).

References
Campbell, I. (2000) The Ostensive Model of developing information needs, PhD Thesis,
University of Glasgow.
Efthimiadis, E.N. (2000) Interactive query expansion: a user-based evaluation in a relevance
feedback environment, Journal of the American Society for Information Science, Vol. 51,
No. 11, pp.9891003.
EU Report (2004a) Does EGovernment pay off?, Available at: http://europa.eu.int/idabc/en/
document/3818/254, November.
EU Report (2004b) eGovernment in the EU in the next decade: the vision and key challenges,
Technical Report EUR 21376 EN, August.
Grimm, S., Motik, B. and Preist, C. (2004) Variance in e-business service discovery, Semantic
Web Services Workshop at ISWC 2004, November.
Haase, P. and Stojanovic, L. (2005) Consistent evolution of OWL ontologies, Proceedings of the
Second European Semantic Web Conference (ESWC 2005), pp.182197.

92

L. Stojanovic, N. Stojanovic and D. Apostolou

Hess, A. and Kushmerick, N. (2003) Automatically attaching semantic metadata to web services,
Proceedings of IJCAI-03 Workshop on Information Integration on the Web (IIWeb-03),
pp.111116.
Klischewski, R. (2004) Semantic Web for e-government a research agenda, AIS SIG SEMIS
Newsletter, Vol. 1, No. 1.
Li, L. and Horrocks, I. (2003) A software framework for matchmaking based on semantic web
technology, Proceedings of the 12th International World Wide Web Conference
(WWW 2003), pp.331339.
Paolucci, M., Kawamura, T., Payne, T.R. and Sycara, K. (2002) Importing the Semantic Web in
UDDI, in Web Services, E-Business and Semantic Web Workshop.
Patil, A., Oundhakar, S., Sheth, A. and Verna, K. (2004) METEOR-S web service annotation
framework, Proceedings 13th International WWW Conference, pp.553562.
Stojanovic, L. (2004) Methods and Tools for Ontology Evolution, PhD Thesis, University of
Karlsruhe, Available at: http://www.ubka.uni-karlsruhe.de/cgi-bin/psview?document=/2004/
wiwi/10&search=/2004/wiwi/10.
Stojanovic, L., Abecker, A., Stojanovic, N. and Studer, R. (2004) On managing changes in the
ontology-based e-government, Proceedings of OTM Confederated International Conferences
(ODBASE 2004), pp.10801097.
Stojanovic, N. (2005) On the Conceptualization of the Query Refinement Task, Library
Management, Emerald Group Publishing, Vol. 26, Nos. 4/5, pp.281294.
Stojanovic, N. and Stojanovic, L. (2004) On modeling cooperative retrieval using an
ontology-based query refinement process, Proceedings of 23rd International Conference on
Conceptual Modeling (ER 2004), pp.434449.

Notes
1

http://www.ontogov.com.
We do not consider dynamic e-government services whose process flow can be composed on the
fly. However, we allow the dynamic binding of e-government services during the execution.
Therefore, we focus on static web services, whose composition is explicitly predefined by the
laws.
3
An ontology OS is defined as a specialisation of an ontology O if it includes the ontology O and
extends its entities either at the conceptual level (e.g. by defining the specialisation of a
concept) or at the instance level (e.g. by instantiating a particular concept). We note that
ontology specialisation is a synonym for ontology reuse or ontology modularisation.
4
The Evolution Ontology is about a meta-ontology that is used as a backbone for creating evolution
logs. It models what changes, why, when, by whom and how are performed in a service
ontology.
5
Ontology consistency in general is defined as a set of conditions that must hold for every ontology
(Stojanovic, 2004).
6
Paradoxically, the query refinement should help a user in resolving an information overload.
7
The classification of e-government services is formalised using LifeEvent ontology that is defined
in Section 3.
2