Change and Version Management in Variability Models for Modular
Ontologies
Melanie Langermeier
1
, Thomas Driessen
1
, Heiner Oberkampf
1, 2
, Peter Rosina
1
and Bernhard Bauer
1
1
Software Methodologies for Distributed Systems, University of Augsburg, Augsburg, Germany
2
Siemens AG, Corporate Technology, Munich, Germany
Keywords:
Variability Model, Modular Ontologies, Version Management, Change Management, Enterprise Architecture
Management.
Abstract:
Modular ontology management tries to overcome the disadvantages of large ontologies regarding reuse and
performance. A possibility for the formalization of the various combinations are variability models, which
originate from the software product line domain. Similar to that domain, knowledge models can then be
individualized for a specific application through selection and exclusion of modules. However, the ontology
repository as well as the requirements of the domain are not stable over time. A process is needed, that
enables knowledge engineers and domain experts to adapt the principles of version and change management
to the domain of modular ontology management. In this paper, we define the existing change scenarios and
provide support for keeping the repository, the variability model and also the configurations consistent using
Semantic Web technologies. The approach is presented with a use case from the enterprise architecture domain
as running example.
1 INTRODUCTION
Knowledge management is an important aspect in or-
ganizations. Typically, different models are used to
capture all relevant aspects of the organization. These
models can extend each other, but can also overlay
in some parts. Furthermore, the semantics of two
models can exclude the use of both in one applica-
tion. The composition of modular knowledge mod-
els to one application ontology is dealt with in the
research area of modular ontology management. A
specific domain ontology is created through selection
of different modules from an ontology repository. To
manage the complexity of the different variants that
exist for the composition of an application ontology,
we proposed in (Langermeier et al., 2013) the use of
variability management (MOVO). This is a technique
from product line engineering, that focuses on a con-
sequent and explicit documentation of the variability
on software artifacts. Such a documentation enables
an individualization of software products while keep-
ing control of the rising complexity of the variants.
Variability models (VM) can also be used for modu-
lar ontologies to formalize the dependencies, that are
annotated in them (e.g. owl:import). Furthermore,
these models also formalize domain independent re-
quirements, for instance, that the domain Ontology A
and Ontology B should not be used together. For the
creation of a configuration, this variability knowledge
can be automatically processed, and therewith sup-
ports the domain expert in creating a consistent con-
figuration. The existing MOVO approach, however,
merely presents a methodology to attain an initial
setup, but does not describe how to handle changes in
the VMs, their configurations and the respective on-
tologies. One major challenge is to keep the overall
system consistent when changes occur. Our aim is to
extend MOVO with concepts and techniques to sup-
port change and version management for modular on-
tology management. This includes concepts, to relate
different versions and configurations to each other,
but also technical support to fulfill consistent version
updates or deletions. In section 2 we summarize the
MOVO approach we want to extend. The foundations
of versioning in ontology management are presented
in section 3. The required change events, to support
in the MOVO extension, are described in section 4
using an enterprise architecture (EA) use case as run-
ning example. The technical realization of the change
management approach is described in section 5) using
OWL 2 and SPARQL. We conclude with an evalua-
tion and discussion of related work.
383
Langermeier M., Driessen T., Oberkampf H., Rosina P. and Bauer B..
Change and Version Management in Variability Models for Modular Ontologies.
DOI: 10.5220/0004953603830390
In Proceedings of the 16th International Conference on Enterprise Information Systems (ICEIS-2014), pages 383-390
ISBN: 978-989-758-028-4
Copyright
c
2014 SCITEPRESS (Science and Technology Publications, Lda.)
Figure 2: A variability model for modeling an enterprise architecture with a possible configuration (Langermeier et al., 2013).
2 VARIABILITY MODELS FOR
MODULAR ONTOLOGY
MANAGEMENT
In software product line engineering, the different
variants of the product are managed through VMs.
Those models allow the explicit and consequent doc-
umentation of the software artifacts’ variability in or-
der to enable reuse in the development process (Chen
et al., 2009). An overview of techniques for variabil-
ity management can be found in (Chen et al., 2009)
and (Sinnema and Deelstra, 2007). One well-known
technique is feature modeling. Features capsule be-
havior, visible to the end user, on a logical layer
(Beuche et al., 2004). Examples for feature model-
ing approaches are (Kang et al., 1990), (Kang et al.,
1998) or (Czarnecki et al., 2004). Modular ontology
management tries to overcome the issues of difficult
reuse and performance challenges in large ontologies.
Therefore, the idea is to develop small ontological
models, called modules, that can be composed to a
bigger application ontology. These small modules en-
able (partial) re-use, more efficient reasoning, easier
maintenance and collaborative development (Spac-
capietra et al., 2005; Stuckenschmidt et al., 2009).
Using OWL 2 provides a standardized set of vocab-
ulary to describe the modules as well as the depen-
dencies between the single modules. These are for
example the relations owl:ontologyIRI, owl:imports
or version information (see section 3) (Motik et al.,
2012). In (Langermeier et al., 2013) we introduced an
approach for the management of the dependencies be-
tween such ontological modules as well as their com-
position to one application ontology. We decided to
use VMs to formalize and reason about the depen-
dencies between modular ontologies. For the cre-
ation of such a variability model, first, an ontology
repository (OntRepo), including all used modular on-
tologies, has to be established. Then the dependen-
cies between these modules are analyzed and formal-
ized in the ontological variability model V M
O
. Using
this model as a basis a knowledge engineer (KE) can
create an integrated variability model V M
I
through
strengthening or extending V M
O
. Finally, the domain
expert (DE) can select the features of V M
I
, that shall
be included. Additional modules, that are necessary
to get a consistent application ontology, will be added
automatically. An overview of this concept with the
used models and their dependencies is given in Fig-
ure 1. Figure 2 is an example of V M
I
for the EA do-
Figure 1: Overview of the models and their dependencies
(Langermeier et al., 2013).
main in order to enable a combined meta model from
two different frameworks. It is modeled using the fea-
ture modeling approach from (Kang et al., 1990). The
model captures not only the ontological restrictions in
the single modules, but also restrictions from the do-
main, modeled by a KE. Each rectangle represents a
feature. The shaded ones indicate a possible configu-
ration for an application ontology.
3 STATE OF THE ART
Versioning in Modular Ontology Management.
One of the main problems using modular ontolo-
gies is to manage changes systematically and being
aware of their impacts. Ontology development is a
continuous process, since ontologies evolve and have
to be adapted to a changing environment and addi-
ICEIS2014-16thInternationalConferenceonEnterpriseInformationSystems
384
tional knowledge. A high modularity intensifies the
problem, since a change in one ontology can affect
various other dependent ontologies (Stuckenschmidt
and Klein, 2007; Flouris et al., 2008). (Stucken-
schmidt and Klein, 2007) characterize different kind
of changes in their work: (1) Changes that do affect
the logical theory and therewith affect the compiled
subsumption relationships and (2) changes that affect
only the syntactic representation or the names of con-
cepts and relations. Furthermore, they state that in
real-word scenarios changes mostly are of the second
type and therefore are harmless changes. Neverthe-
less, also the smaller kind of harmful changes require
methods to deal with since they can have an impact
on other modules. Therewith, we do not focus on the
effects of changes within an ontology or the effects on
other ones in this paper. Merely, we address the prob-
lem to find that places in an organization where an on-
tology, that has or should be changed, is in use. When
doing changes to an ontology, this will always result
in a new version. A versioning method is required,
which will deal with the different versions of an ontol-
ogy and the effects of changes to them (Klein, 2001).
A method supporting ontology versions has to keep
track of dependencies between them and provide ac-
cess to ontologies, (prior) versions of ontologies and
their respective usages (Flouris et al., 2008).
Version Information for Ontologies. To repre-
sent version information, OWL 2 provides different
annotation properties (Motik et al., 2012). First,
owl:ontologyIRI can be used to identify an ontol-
ogy. If an ontology IRI is specified, one can addi-
tionally declare an owl:versionIRI. Ontologies with
the same ontology IRI but different version IRIs be-
long to one ontology series. In a series there is ex-
actly one current ontology which should be accessible
through the ontology IRI. Using the owl:priorVersion
property one can relate an ontology to its previ-
ous versions. Information about compatibility be-
tween versions of an ontology can be represented by
owl:backwardCompatible or owl:incompatibleWith.
There is no clear definition of owl:incompatibleWith
within the specification. Obviously two ontology ver-
sions are incompatible, if they contain contradicting
statements
1
, but one could also regard two versions as
incompatible, if not all concepts of the prior version
have the same semantics in the new version. General
version information, like a version number, a date or
other information can be added using the annotation
property owl:versionInfo.
1
http://www.w3.org/TR/owl-guide/.
4 EXTENDING MOVO FOR
VERSION AND CHANGE
MANAGEMENT
The four basic functions of persistent storage, create,
read, update and delete, are called CRUD in the do-
main of computer science. (Langermeier et al., 2013)
shows how to read from the OntRepo and VMs, in
the following the create, update and delete (CUD)
functions are shown. In the MOVO approach intro-
duced in section 2 four results are established: (1) an
ontology repository (OntRepo), containing the rele-
vant ontological modules for the organization, (2) an
ontological variability model V M
O
, formalizing the
dependencies between the modules, (3) an integrated
variability model V M
I
, formalizing additional depen-
dencies according to a specific domain under consid-
eration of V M
O
and (4) a specific configuration of
modules, which is conform to V M
I
and serves as ap-
plication ontology. For each function, the existing de-
pendencies between the OntRepo, the different VMs
and their configurations are shown, as well as how
these dependencies can be utilized to end up with a
consistent overall system, after performing one of the
CUD functions. Typically, such a repository stays not
stable over time, it changes. To be able to manage
these changes and ensure the correctness of the exist-
ing application ontologies, we enhance MOVO with
concepts for the classification and connection of dif-
ferent V M
I
s as well as the configurations based on
those V M
I
s. Figure 3 shows MOVO extended with
concepts for versioning and tracing. To be able to
Figure 3: Extend concept of MOVO with versioning.
trace configurations to the V M
I
they are built on, a
configuration of/has configuration relationship is re-
quired. To connect different V MI
I
s, a successor/pre-
decessor relationship is introduced. As proposed for
ontologies by the OMG, we also use backward com-
patible and incompatible relationships between V M
I
s.
ChangeandVersionManagementinVariabilityModelsforModularOntologies
385
To mark obsolete V M
I
s, the label deprecated is in-
troduced. The technical details about the new rela-
tionships are explained in section 5. In the following
the different possible change events are illustrated and
their effects are shown using the Enterprise Architec-
ture use case of (Langermeier et al., 2013) as running
example (see figure 2).
4.1 Changes in the Ontology Repository
Create Ontology. In order to model the distribution
and communication path of the software applications
in an organization, the KE wants to add an ontology
module about communication and distribution. This
new module has two important dependencies to
existing modules in the repository. The new ontology
is added to the OntRepo in order to be further pro-
cessed. The V M
O
data set is updated automatically
using the set-up steps described in section 5.2. This
ensures that the OntRepo and V M
O
are always
consistent to each other. In the use case, the new
ontology will not be included in V M
I
at that time. It
is the responsibility of the KE to decide, whether the
new ontology should be included in an existing V M
I
.
In that case, the changes to V M
I
have to be published
as a new version (for more details see Update V M
I
).
Update Ontology. When updating an ontology,
only the ontology’s version information is consid-
ered and internal ones, i.e., semantic or syntactic
changes are not interpreted. That means, when
an updated version of an ontology is added to
the OntRepo, the V M
O
is updated differently,
according to the ontology’s comprising OWL
annotations (owl:backwardCompatibleWith and
owl:incompatibleWith). If only an owl:priorVersion
annotation is made, without further information
about compatibility, this is treated as an incompatible
update. The first and simple case is the annotation
of owl:backwardCompatibleWith: the ontology can
be simply included in the OntRepo. The V M
O
must not be changed, since an ontology’s IRI always
represents its newest version. If applicable, additional
entailed dependencies of this newer version are also
checked, adapted and added to V M
O
. The second
case implies, that the ontology is annotated with
owl:incompatibleWith. In this case, the representing
features of the old and new version get translated
to an XOR group in the feature model and both are
kept in V M
O
. For example The TOGAF Business
module is updated to version 2.0, while keeping it
backward compatible with the prior version 1.0. The
Iteraplan Business module is also updated to version
2.0, but will then be incompatible with the prior
version. Figure 4 visualizes the changes in the EA
V M
O
through the update of Iteraplan Business. To
Figure 4: Visualization of the publication of a new version,
incompatible to the old one in V M
O
.
keep the Iteraplan Business feature mandatory, an
empty feature is created which is then decomposed
using the XOR group. Afterwards, the KE has to
determine if the changes made to the newer version
have an impact on the already existent mappings.
When necessary, they have to be adapted manually,
according to the changes made to the ontology. If the
KE wants to use the new versions in the V M
I
, this has
to be done manually via an update (see Update V M
I
).
Delete Ontology. Although deletion of an existent
and used ontology is very unlikely to happen, the fol-
lowing paragraph describes how to keep the overall
system consistent after such an event. This change
event is illustrated through deletion of the TOGAF
Motivation module as well as the old versions of TO-
GAF Business and Iteraplan Business. In contrast to
the former two operations the OntRepo is unaffected
for the moment. The to be deleted ontology is marked
as deprecated. This implies, that the V M
I
s containing
that ontology are marked as deprecated too. Since the
TOGAF Motivation module is not used in any V M
I
,
we are able to safely remove this module from the On-
tRepo and V M
O
. Whereas the modules TOGAF Busi-
ness and TOGAF Iteraplan are included in an V M
I
,
we cannot remove these modules. To be able to exe-
cute the deletion, the KE has to delete the respective
V M
I
s (see Delete V M
I
).
4.2 Changes in the Set of V M
I
s
Create V M
I
. The creation of a new V M
I
does not
have any impact, neither on the OntRepo nor on
existing configurations.
Update V M
I
. A V M
I
can only be updated via the
creation of a new version. This new version has
ICEIS2014-16thInternationalConferenceonEnterpriseInformationSystems
386
to be linked via the successor-relationship to its
prior one. Using the labels incompatible with and
backward compatible more semantic information
about the update can be made. In the new version
of V M
I
for EA (V2) only the newer version of the
Iteraplan Business module should be included. Due
to the incompatibility of the module versions also the
two V M
I
versions are incompatible. Whenever an
V M
I
is updated, all DEs of affected configurations
are determined via the link between the V M
I
and its
configurations. If the DE updates the configuration
to the new version of the V M
I
, the link between the
configuration has to be updated accordingly.
Delete V M
I
. A V M
I
can only be deleted, if there is
no configuration using it. Following, before deleting,
the V M
I
has to be marked as deprecated. If the DE
decides to switch his application to V M
I
V2, there is
no configuration using V M
I
V1 and it can thus be re-
moved. Every time a V M
I
is removed by a KE, this
V M
I
has to be checked for a deprecated ontology. If
such an ontology was found, and no other VM
I
is us-
ing this ontology, it can be removed safely from the
OntRepo and V M
O
. When deleting V M
I
V1, we get
informed, that now the deprecated modules TOGAF
Business V1 und Iterplan Business V1 are no longer
in use in any configuration or V M
I
, and can be re-
moved safely.
4.3 Events Concerning the
Configurations
Create/Update Configuration. Every time a DE
creates or updates a configuration, the link to the
respective V M
I
(movo:configurationOf ) will be
created/updated.
Delete Configuration. A configuration can be
deleted by a DE at any time. If the corresponding
V M
I
is marked as deprecated and has no more pend-
ing configurations left, this V M
I
can be deleted too.
In our use case, after the deletion of the old configu-
ration, we get informed that now the deprecated VM
I
V1 is no longer used by any configuration. We are
now able to remove V M
I
V1.
5 REALIZATION
In the following the extended version of the Variabil-
ity Model Ontology (VMO), the storage of ontology
versions in the OntRepo, the representation of them as
features in the ontological variability model and the
change management are described.
5.1 Variability Model Ontology
The initial version of the VMO is described in
(Langermeier et al., 2013). The main classes
of the V M
O
are vmo:Feature
2
, owl:Ontology,
vmo:Composition, vmo:Alternative_Composition and
vmo:Or_Composition. We added several new object
properties to the V M
O
(see figure 5). For instance,
Figure 5: Object properties of the Variability Model Ontol-
ogy (VMO). New properties are in bold.
we defined an object property vmo:isRealizedIn to re-
late features with ontology IRIs of the OntRepo (see
Figure 8). Domain and Range of this property is
vmo:Feature and owl:Ontology respectively. Further,
we added a class vmo:VMI for the integrated Variabil-
ity Models (V M
I
s). A V M
I
is an instance of vmo:VMI
and is represented through a named graph, contain-
ing features and relations between them. Named
graphs are a good way to group triples in a triple
store and further the graph URI (i.e. the URI of
the V M
I
instance) can be used to relate the V M
I
in-
stance to its dependent configurations: A configura-
tion is a valid selection of features from one V M
I
.
Since each feature represents an ontology, a con-
figuration represents indirectly also a set of ontolo-
gies. The two properties vmo:configurationOf and
vmo:hasConfiguration are used to connect a con-
figuration with the corresponding V M
I
. Selected
features are related to the configuration with the
property vmo:selectedFeature. The object prop-
erty vmo:isRealizedIn is used to relate vmo:Features
with the respective ontology the feature is represent-
ing. Additionally a transitive property vmo:successor
(with inverse vmo:predecessor) is used to intercon-
nect different versions of V M
I
s in the right order. Do-
main and range of both properties is vmo:VMI. Differ-
ent versions of V M
I
s are backward compatible (and
linked with vmo:backwardCompatibleWith property),
2
We write vmo:Feature instead of <http://www.ds-
lab.org/ontologies/movo-ontology.owl#Feature>.
ChangeandVersionManagementinVariabilityModelsforModularOntologies
387
if all configurations created by the older version can
still be created using the newer version. Otherwise
they are linked with vmo:incompatibleWith. The
property owl:deprecated is generally used for mark-
ing IRIs as deprecated. We use this annotation prop-
erty to flag outdated and thereby obsolete ontolo-
gies, which comes into play when ontologies shall be
deleted. A similar property (vmo:deprecated) is used
to flag V M
I
s, configurations or features as as depre-
cated.
5.2 Set-up of OntRepo and VMO
For the technical realization of our approach we set
up an Apache Jena Fuseki triple store with two data
sets (see figure 6). The first data set (ontrepo) is
Figure 6: The triple store contains two data sets. One for the
ontology repository and another for the variability models.
A SPARQL Query Repository is used to access and modify
data.
the ontology repository, the second (vmo) is used
to store the Variability Model Ontology, the features
with their dependencies and the V M
I
s. Since the on-
tology repository contains several versions of one on-
tology we separate them using named graphs. This is
necessary since concepts of different versions share
the same namespace. The named graphs reuse the
version IRIs of the ontology versions or the ontology
IRI if no version IRI is defined. Ontology annotation
properties such as version IRIs or import statements
are additionally stored in the default graph. These
relations are shown in Figure 7. In the next step,
we automatically create features in the vmo dataset
for all ontology IRIs in the default graph of the On-
tRepo data set. This is done using SPARQL UP-
DATE queries from the query repository. If there
are incompatible versions of one ontology, we ad-
ditionally create one vmo:Alternative_Composition
Figure 7: Relations between ontology IRIs in the OntRepo
default graph.
and relate it to the version IRIs and the ontology
IRI through vmo:consistsofFeatures property. Since
vmo:consistsOfFeatures has range vmo:Feature these
versions become also vmo:Feature in the vmo data
set. An example of an alternative composition with
relations to the IRIs is shown in figure 8. Within
Figure 8: An alternative composition in V M
O
and relations
to ontology IRIs from the OntRepo.
the creation process of V M
I
the KE can then decide
whether to use the feature for the ontology (which
will result in the use of the most current version) or
the composition rule in order to allow the use of one
of the different versions.
5.3 Changes in the Ontology Repository
If a new ontology is added to the ontology repository,
the initial set-up steps for integration into OntRepo
and V M
O
as described in section 5.2 are repeated.
If a new version of an existing ontology is added to
the repository, we distinguish two cases: If the new
version is backward compatible, we simply repeat the
set-up steps. Since the corresponding feature points to
the ontology IRI and not to the version IRI, we do not
have to change the V M
O
. However configurations us-
ing the respective features have to be recreated. If the
new version is incompatible with the prior version we
need to create a new alternative composition in V M
O
or adapt an existing one in adding a new feature for
the new version. If one ontology should be deleted
from the ontology repository,first the IRI of the ver-
sion or ontology has to be flagged as deprecated us-
ing the owl:deprecated property as mentioned above.
Features in the V M
O
dataset, which are related to a
deprecated ontology via the vmo:isRealizedIn prop-
erty, as well as corresponding V M
I
s are set as depre-
cated using vmo:deprecated property too. The feature
and corresponding ontology are deleted after the last
ICEIS2014-16thInternationalConferenceonEnterpriseInformationSystems
388
V M
I
graph referencing the feature is dropped from the
V M
O
dataset.
5.4 Changes in V M
I
A backward compatible new version of a V M
I
simply
obtains all references from configurations to the old
version. The named graph of the old VM
I
is dropped
afterwards. Both operations are automatically per-
formed using SPARQL UPDATE queries. Configu-
rations are not linked to the new V M
I
, if it is incom-
patible to the former version. In the case of ontology
updates in the OntRepo all V M
I
s containing features
with references to the respective ontology need to be
updated.
6 EVALUATION
Comparing our versioning method with the issues
stated in section 3, we support most of them. We are
able to manage different versions of the same ontol-
ogy in one data set and also manage the dependencies
between those versions. We consider the dependen-
cies between different ontologies as well as their us-
age in specific applications. Therewith, our method
ensures, that the information about which ontology
version is used in which application is always correct
and hinders an automatic update to newer ontology
versions that may cause inconsistencies. Older ver-
sions of an ontology are kept in the data set while
marking them as deprecated. If no other ontology or
application is using them, they can be automatically
removed. Additionally, we evaluated the method in a
use case (see running example in section 4). There-
fore, we reused the established V M
I
(Figure 2) from
(Langermeier et al., 2013) and executed change oper-
ations. This VM is created for the Enterprise Archi-
tecture (EA) domain in order to be able to combine
different meta models and mix and match them ac-
cording to the requirements of a specific organization.
The different versions of the models (current, depre-
cated, deleted) with their dependencies, that exist af-
ter executing all the change operations, are illustrated
in figure 9
3
.
3
The test set for the evaluation with the data sets,
queries, scripts and a documentation is published in
http://megastore.uni-augsburg.de/get/HAth0VS7qw/. The
test set was built upon the results from (Langermeier et al.,
2013).
Figure 9: Concept applied to EA use case after executing
the change operations.
7 RELATED WORK
Ontology Evolution. As stated in (Stojanovic et al.,
2002), most of the research in the ontology engineer-
ing field is done on construction issues. However,
evolution of ontologies requires different approaches.
Our work was focused on the evolution of multiple,
modular ontologies on an abstract feature level,
whereas most of the related work copes with single
ontologies and their evolution over time. Examples
can be found manifold, like in (Stojanovic et al.,
2002). In this work, the authors present a 6-phase
methodology in order to keep an ontology consistent
over time. The key step of this methodology is to
differentiate between syntactic and semantic changes.
Another approach is presented by (Plessers et al.,
2007), where the authors created an own framework
for detecting changes, made to OWL DL ontologies.
Using a predefined set of change definitions, they are
able to provide different overviews of changes made
to an ontology, depending on the user’s role.
Ontology Repositories. Ontology repositories store
either domain specific or independent ontologies, pro-
vide users with browse and search interfaces, and
can offer other functionality like reviewing, annotat-
ing or editing and mapping ontologies. The ontolo-
gies’ metadata, for instance, author, version, date or
scope, is a valuable instrument for these functionali-
ties. The mapping metadata can even be used to relate
concepts from different ontologies (Noy et al., 2008).
Generally, there are two different types of ontology
repositories: the ones that are automatically crawled
(Swoogle, Watson etc.) and the ones where the on-
tologies are submitted manually by users (Noy et al.,
2008). The initialization and extension of our pro-
posed repository is a manual effort. It is still rudimen-
tary and does not provide a user interface. Nonethe-
ChangeandVersionManagementinVariabilityModelsforModularOntologies
389
less, we use the meta data provided with the ontolo-
gies in order to find mapping and version information.
8 CONCLUSION
In this paper we proposed a method for change and
version management for MOVO. By reusing the well
known CRUD functions, we strongly reduced the
complexity for change management. We show the
impact of those events for the OntRepo, its VM
O
, the
different V M
I
s and their configurations. Using tech-
nologies from the Semantic Web technology stack,
we implemented the method to calculate the impacts
automatically. Some parts still require the decision
of a KE or a DE. Applying this method we enable
MOVO to be used by long-term systems without los-
ing manageability due to increasing complexity. Fu-
ture work has to be done to enable our approach han-
dling the changes made within an ontology and es-
pecially consider the semantic impacts these mod-
ifications have on other ontologies. We also want
to research more expressive approaches for defin-
ing relations between ontologies, e.g., E -Connections
(Cuenca Grau et al., 2009), Package Based Descrip-
tion Logics (P-DL) (Bao et al., 2006), Distributed De-
scription Logics (DLL) (Borgida and Serafini, 2003)
or the Interface-based modular ontology Formalism
(IBF) (Ensan, 2010). Finally a prototype tool imple-
mentation has to be done with a first user interface.
REFERENCES
Bao, J., Caragea, D., and Honavar, V. G. (2006). Modular
ontologies a formal investigation of semantics and
expressivity. In Proceedings of the First Asian con-
ference on The Semantic Web, ASWC’06, pages 616–
631, Berlin, Heidelberg. Springer-Verlag.
Beuche, D., Papajewski, H., and Schröder-Preikschat, W.
(2004). Variability management with feature models.
Science of Computer Programming, 53(3).
Borgida, A. and Serafini, L. (2003). Distributed Description
Logics: Assimilating Information from Peer Sources.
Journal on Data Semantics, 1:153–184.
Chen, L., Babar, M. A., and Ali, N. (2009). Variability
management in software product lines: a systematic
review. In Proceedings of the 13th International Soft-
ware Product Line Conference, SPLC’09, pages 81–
90.
Cuenca Grau, B., Parsia, B., and Sirin, E. (2009). Modu-
lar ontologies. chapter Ontology Integration Using E-
Connections, pages 293–320. Springer-Verlag, Berlin,
Heidelberg.
Czarnecki, K., Helsen, S., and Eisenecker, U. (2004).
Staged Configuration Using Feature Models. In Nord,
R. L., editor, Software Product Lines, volume 3154 of
Lecture Notes in Computer Science, pages 266–283.
Springer Berlin Heidelberg.
Ensan, F. (2010). Semantic Interface-Based Modular On-
tology Framework. PhD thesis, University of New
Brunswick.
Flouris, G., Manakanatas, D., Kondylakis, H., Plexousakis,
D., and Antoniou, G. (2008). Ontology change: clas-
sification and survey. The Knowledge Engineering Re-
view, 23(2).
Kang, K., Kim, S., Lee, J., Kim, K., Shin, E., and Huh,
M. (1998). FORM: A feature-oriented reuse method
with domain-specific reference architectures. Annals
of Software Engineering, 5(1):143–168.
Kang, K. C., Cohen, S. G., Hess, J. A., Novak, W. E.,
and Peterson, A. S. (1990). Feature-Oriented Domain
Analysis (FODA) Feasibility Study. Technical report,
Carnegie-Mellon University Software Engineering In-
stitute.
Klein, M. (2001). Combining and relating ontolgies: an
analysis of problems and solutions. In IJCAI 2001
Workshop on ontologies and information sharing,
pages 53–62.
Langermeier, M., Rosina, P., Oberkampf, H., Driessen, T.,
and Bauer, B. (2013). Management of Variability
in Modular Ontology Development. In International
Workshop on Semantic Web Enabled Software Engi-
neering, Berlin.
Motik, B., Patel-Schneider, P. F., and Parsia, B. (2012).
OWL 2 Web Ontology Language, Structural Specifi-
cation and Functional-Style Syntax (Second Edition).
W3C Recommendation.
Noy, N. F., Griffith, N., and Musen, M. A. (2008). Collect-
ing community-based mappings in an ontology repos-
itory. The Semantic Web-ISWC 2008, pages 371–386.
Plessers, P., De Troyer, O., and Casteleyn, S. (2007). Under-
standing ontology evolution: A change detection ap-
proach. Web Semantics: Science, Services and Agents
on the World Wide Web, 5(1):39–49.
Sinnema, M. and Deelstra, S. (2007). Classifying variabil-
ity modeling techniques. Journal of Information and
Software Technology, 49(7).
Spaccapietra, S., Menken, M., Stuckenschmidt, H., Wache,
H., Serafini, L., and Tamilin, A. (2005). Report on
Modularization of Ontologies. Knowledge Web Con-
sortium, (D2.1.3.1).
Stojanovic, L., Maedche, A., Motik, B., and Stojanovic, N.
(2002). User-driven ontology evolution management.
In Knowledge engineering and knowledge manage-
ment: ontologies and the semantic web, pages 285–
300. Springer.
Stuckenschmidt, H. and Klein, M. (2007). Reasoning and
change management in modular ontologies. Data &
Knowledge Engineering, (63):200–233.
Stuckenschmidt, H., Parent, C., and Spaccapietra, S., ed-
itors (2009). Modular Ontologies: Concepts, Theo-
ries and Techniques for Knowledge Modularization,
volume 5445 of Lecture Notes in Computer Science.
Springer, Berlin.
ICEIS2014-16thInternationalConferenceonEnterpriseInformationSystems
390