Carsten Felden
Technical University Bergakademie Freiberg, Lessingstraße 45,09596 Freiberg,Germany
Daniel Kilimann
Mercator School of Management, University Duisburg-Essen,Campus Duisburg,47057 Duisburg,Germany
Keywords: Management Information Systems, Ontologies, Ontological Engineering.
Abstract: The consideration of integrated structured and unstructured data in management information systems
requires a new kind of metadata management. Ontologies constitute a possibility to solve the resulting
problems. Process models describe the development of ontologies which can be utilised in the context of
management information systems, are discussed.
The penetration of companies with information
systems constitutes the basis of information supply
for decision makers by means of structured and
unstructured data integrated from internal and
external sources. More and more heterogeneous data
are available and this leads to the often quoted
information flooding. Furthermore, heterogeneous
data formats lead to the problem that related data
cannot be found and shown to the user. Semantic
annotation of natural language documents and the
integration of domain ontologies can enable
semantic inquiries.
This paper presents an idea of melting data
dictionary data with ontologies in favour of queries
in an integrated database. First of all, we introduce
business intelligence (BI) systems which are using
such an integrated database in order to gain an
impression of available data for the reader. Chapter
three addresses the issue of ontologies and their
possible application fields. A restraint on the
ontology usage is their development as well as their
maintenance. Therefore chapter four discusses
current approaches of ontology development.
Chapter five recapitulates the results and gives an
Up to 90 percent of the information in a company
are not available in a machine processable format,
e.g. structured data, but as unstructured, non-
machine processable, data. These kinds of data are
generally natural language documents (Kantardzic,
2003). Due to this reason, there is a considerable
potential which can be managed by an adequate
handling of information flooding. Data in BI-
systems derive from different heterogeneous
sources. They have to be differentiated between
intra-corporate and external sources. The first ones
are operational application systems. The operational
system environment is heterogeneous, because these
systems were normally developed isolated from each
other. Operational applications are using different
data structures and formats. External data can be for
instance purchased data streams from news services
like Reuters or the result of queries sent to search
engines like Google. Appropriate systems are
required in order to retrieve all stored information of
the central database.
The integration process of internal and external
data is comparable. During the transformation, data
are syntactically and semantically adjusted. The
Felden C. and Kilimann D. (2006).
In Proceedings of the Eighth International Conference on Enter prise Information Systems - DISI, pages 306-309
DOI: 10.5220/0002453603060309
syntactic data transformation is the necessary
conversion of formats into a uniform standard.
Semantic transformation deals with clearing up
textual senseless field contents, the decomposition of
semantic overburdened fields and the elimination of
synonyms and homonyms. The results of the
transformation process are data structures which
correspond to the design of the database included in
the Business Intelligence system.
In the scope of computer science an ontology is
formally a defined system of concepts. This paper is
based on the definition of Studer: “An ontology is a
formal, explicit specification of a shared
conceptualisation.” (Studer et al., 1998)
Conceptualisation corresponds to an abstract
model of a domain which identifies the relevant
concepts and their relationships. Explicit means that
the used concepts are unique and their usage is
formally confined. Formal refers to the fact that
ontologies should be machine-readable. Shared
indicates that an ontology is accepted by a group of
people and used corporately.
Important components of a data warehouse are
metadata. They link the operational information
systems and the data warehouse of a BI system.
Metadata are located in a directory which enables
analysts to discover data in a meta database system.
This directory is called data dictionary or repository
(Froeschl, 1997). Metadata consist of all information
which simplify development, maintenance and
administration of a data warehouse system as well as
enable the acquisition of information for the data
warehouse (Bauer and Günzel, 2001). They explain
the transformations during the data integration
process. Furthermore, they characterize the
algorithms operating in the data warehouse so that
the result is the linkage between the aggregation
processes and the subject orientation of the entire
database (Inmon and Hackathorn, 1994).
More precisely, a central requirement of a data
dictionary is the documentation of the data fields
and database structures including data origin, data
validation, data definition, possible influencing
variables, details about the acquisition of data, and
links to other information (Wertz, 1986). It can be
stated that a data dictionary concentrates on
structured data. But, as mentioned above, a BI
system also covers unstructured data. Natural
language documents are characterized by a
confusing variety of terms. According to this
situation, metadata have to consider synonyms and
multilingual terms. Ontologies offer corresponding
assistance in this context. Their major requirement is
to make such information machine-processable and
to simplify accessing data.
The data dictionary provides an appropriate basis
to construct an ontology, because it ensures the
unambiguousness of the used terms within the
database and contains the necessary metadata.
Because of this it is a suitable foundation to identify
appropriate concepts und their relationships. But, the
modelling process has to be executed manually.
Models are especially important in order to
recognise and eliminate possible restraints during
the development of the ontology and to simplify the
maintenance procedure.
Although ontology development is comparable with
software development life cycles, special
requirements of ontologies have to be kept in mind.
In the recent years, numerous suggestions were
made how to develop an ontology (Staab et al.,
4.1 Ontology Development
The METHONTOLOGY approach was published
by Fernandez-Lopez, Gomez-Perez and Juristo in
1997 (Fernandez-Lopez et al., 1999).
METHONTOLOGY is a comprehensive ontology
development methodology according to the IEEE-
norm in the fields of software development and
knowledge management. The activities of the
ontology development process are divided into three
categories: project management activities,
development activities and supporting activities.
Development activities describe the procedure of
ontology construction in detail. Project management
activities include planning, control, and quality
assurance. Both have to be distinguished from the
accompanying supporting activities. These activities
are divided into knowledge acquisition, integration,
evaluation, documentation, and configuration
management (Corcho et al., 2003).
The On-To-Knowledge project is concentrated on
the design of an ontology based knowledge
management system. The On-To-Knowledge
procedure model consists of the following phases:
feasibility study, kickoff-phase, refinement,
evaluation, and maintenance (Sure, 2002). The
feasibility study identifies chances and risks and
analyses the primary application areas. The results
of this phase are the basis for the kickoff-phase. The
created application specification contains the
domain, the objective, design directives, available
resources, and potential users. Subsequent
competence questions are formulated in order to
collect domain specific terms in an informal manner.
The main focus of the TOVE (Toronto Virtual
Enterprise) methodology, created by Grüninger and
Fox, is to provide a series of competence questions
(Grüninger and Fox, 1995). Questions on the
problems that have to be solved are formulated and
should be answered afterwards by the ontology.
They are used in order to build the concept hierarchy
and to evaluate the ontology.
The SENSUS approach was introduced by
Swartout (Swartout et al., 1996). The initial point is
the SENSUS ontology itself. It represents an
extensive ontology including 70,000 domain
independent concepts. Representative concepts of
this domain are selected and manually linked with
the SENSUS ontology. Afterwards, all concepts are
inserted which are located directly at the path from
the specific terms to the root. Further concepts are
included manually. The remaining SENSUS
concepts are discarded as irrelevant.
The KACTUS approach was developed with the
scope of the Esprit-project. It postulates already
existing ontologies which are reused or customized
in order to create a new one (HCS, 1996). First of all
the applications, thus the relevant concepts and
objectives, are specified. A new ontology is
developed by adjusting and refining the already
existing top-level or reusable ontologies.
4.2 Critical Review
Following the framework of Gómez-Pérez,
Fernández-López, and Corcho, two different kinds
of criteria are used to evaluate the shown approaches
(Gómez-Pérez et al., 2004). The first criteria type
follows the IEEE-standard using the same criteria as
in the field of software development (Fernández-
López et al., 1999). According to the IEEE-norm
1074-1997, criteria can be classified into three
categories: ontology management activities,
development oriented activities, and accompanying
activities (IEEE, 1998). The ontology management
activities constitute the tasks and functions of the
project management within the development
process. The development oriented activities splits
into predevelopment, development, and
postdevelopment. The accompanying activities as
mentioned above support the development process
and are executed parallel to it.
The activities described in the Grüninger and
Fox methodology concern pre- and postdevelopment
as well as ontology and configuration management.
There are fewer activities described in the KACTUS
approach. Furthermore the accompanying activities
are missing. METHONTOLOGY refers to almost
every activity, but the descriptions differ in their
level of detail. These which are part of the
predevelopment and the implementation are missing.
The On-To-Knowledge methodology covers the
entire spectrum of suggested activities including the
ones for the predevelopment process. Finally,
SENSUS does not include the phases
conceptualisation and formalisation.
Further criteria are life cycle, application
dependence, and the usage of a core ontology. There
are two options an incremental life cycle or
evolutionary prototyping. There are three different
specifications concerning the next criterion:
application dependent, application independent and
semi-application dependent. Reusability of existing
ontologies enables an efficient handling of available
knowledge. The usage of a data dictionary as a basis
of ontology development fits in this context.
The approach of Grüninger and Fox as well as
On-To-Knowledge supports both possible life
cycles. In contrast, concerning KACTUS and
METHONTOLOGY evolutionary prototypes are
recommended. The SENSUS approach provides no
life cycle at all.
The Grüninger and Fox methodology just as the
SENSUS approach are characterized as semi-
application dependent. The KACTUS and On-to-
Knowledge approaches are application dependent by
definition. In contrast to this, METHONTOLOGY is
application independent.
Finally reuse of existing ontologies is discussed.
Reusability is not part of the Grüninger and Fox
methodology. According to the KACTUS approach
new ontologies are developed by reusing or
adjusting existing ones. Analogue to this, the
SENSUS ontology is used as a basis for constructing
the designated ontology. A reusable ontology is not
mentioned explicitly in METHONTOLOGY as well
as in On-To-Knowledge, but the idea is taken into
account in both approaches.
4.3 Application of Ontologies in BI
The following figure shows the assignment between
a data-dictionary entry and an ontology entry.
Dimension Product = {product entry}
product entry = Product_ID + Product_Name +
Type + (Hour) + (Day) + (Week) + (Month) +
(Quarter) + Start_Year + End_Year
< daml: ObjectProperty rdf: ID = “traded at” >
< rdfs: domain rdf: resource = “{product entry}” >
< rdfs: range rdf: resource = ”APX”/ >
< /daml: ObjectProperty >
< daml: Class rdf: ID = “Amsterdam Power Exchange” >
< daml: sameClassAs rdf: resource = “’#APX”/ >
</ daml: Class >
Figure 1: Data dictionary with connected ontology.
The data dictionary entry in the superior part of
the figure describes the product dimension of a
database as part of a BI system. The lower part
shows an ontology entry. The line
< rdfs: domain rdf:
resource = “{product entry}” >
references to the respective
data dictionary entry. This means an enhancement of
the ontology model and creates a connection
between the technical and semantic product
description. Users can benefit from an integrated
view on structured and unstructured data based on
the above described connection.
BI systems provide their users access to structured
and unstructured data. The problem of an integrated
metadata management is not solved, yet. Ontologies
are a presently discussed proposal. An existing data
dictionary administrating structured data should be
enriched with functionalities of an ontology in order
to be able to handle unstructured data as well. The
development of an ontology based on an existing
data dictionary requires a large manual effort. Due to
this reason, common ontology development
procedure models are discussed in this paper. A
terminal decision is not yet possible, because models
are based on different assumptions and aims. In
addition, new approaches have to be recommended.
Furthermore, it has to be clarified in the future, if
semi-automatic methods can be integrated into a
standardized ontology development process.
Bauer, A. and Günzel, H., 2001. Data-Warehouse-
Systeme. Architektur, Entwicklung, Anwendung,
dpunkt.verlag. Heidelberg.
Corcho, Ó.; Fernández-López, M. and Gómez-Pérez, A.,
2003. Methodologies, Tools and Languages for
Building Ontologies. Where is their meeting point? In
IEEE Transactions on Data and Knowledge
Engineering 46 (2003) 1, p. 41-64. id=864179.2003,
accessed 2005-08-20.
Fernández-López, M., Gómez-Pérez, A., Pazos-Sierra, A.
and Pazos-Sierra, J., 1999. Building a Chemical
Ontology Using Methontology and the Ontology
Design Environment. In: IEEE Intelligent System 14
(1999) 1, p. 37-46.
Froeschl, K. A., 1997. Metadata Management in
Statistical Information Processing - A Unified
Framework for Metadata-Based Processing of
Statistical Data Aggregates, Springer. Wien, New
Gómez-Pérez, A., Fernández-López, M., Corcho, O.,
2004. Ontological Engineering, Springer, London et
Grüninger, M. and Fox, M. S., 1995. Methodology for the
Design and Evaluation of Ontologies. In Proceedings
of IJCAI-95 Workshop on Basic Ontological Issues in
Knowledge Sharing. Montreal, Canada.
Human-Computer Studies Lab, 1996. The KACTUS
Booklet version 1.0. Esprit Project 8145 KACTUS,
Reports.html, accessed 2005-08-20.
Institute of Electrical and Electronics Engineers
, 1998.
IEEE Standard for Developing Software Life Cycle
Processes. IEEE Std 1074-1997.
er=16018. 1998, accessed 2005-08-21.
Inmon, W. H. and Hackathorn R. D., 1994. Using the data
warehouse. Wiley, New York et al..
Kantardzic, M., 2003. Data Mining. Concepts, Models,
Methods, and Algorithms. IEEE Press, Piscataway.
Staab, S., Schnurr H. P., Studer, Rudi and Sure, York,
2001. Knowledge Processes and Ontologies. In IEEE
Intelligent Systems 16 (2001) 1, p. 26-34.
Studer, R., Benjamins, R. V., Fensel, Dieter, 1998.
Knowledge Engineering. Principles and Methods. In
IEEE Transactions on Data and Knowledge
Engineering 25 (1998) 1-2, p. 161-197.
Sure, Y.: On-To-Knowledge. Ontology-based Knowledge
Management Tools and their Application. In KI 14
(2002) 1, p. 35-37.
Wertz, C. J., 1986. The Data Dictionary - Concepts and
Uses. North-Holland, Amsterdam.