APPLYING TERMINOLOGICAL METHODS AND
DESCRIPTION LOGIC FOR CREATING AND IMPLEMENTING
AN ONTOLOGY ON INHIBITION
Sine Zambach
CBIT, Roskilde University, Universitetsvej 1, Roskilde, Denmark
Bodil Nistrup Madsen
ISV, Copenhagen Business School, Dalgas Have 15, Frederiksberg, Denmark
Keywords: Ontology modeling, Formal ontologies, Terminological methods, Description Logic.
Abstract: By applying formal terminological methods to model an ontology within the domain of enzyme inhibition,
we aim to clarify concepts and to obtain consistency. Additionally, we propose a procedure for
implementing this ontology in OWL with the aim of obtaining a strict structure which can form the basis for
reasoning and further processing, and we compare a semi-formal terminological concept modeling approach
with a formal Description Logic approach in OWL-DL.
1 INTRODUCTION
Much salient work is put into formalizing
biomedical ontologies using Description Logic,
usually with the purpose of checking consistency, cf.
for example SNOMED CT (Sterns et al., 2001).
Description Logic allows for a formal description
via a wide range of roles, classes and instances, and
it has the possibility of expressing a number of
logical descriptions related to each class (Baader et
al., 2003). However, this possibility of DL can be
inconvenient when a minimization of the number of
conditions describing each concept is desired, and
we therefore argue that some modeling restrictions
could be useful.
Terminological concept modeling uses delimiting
characteristics to clarify how subordinate concepts
of the same superordinate concept differ from each
other (ISO 704, 2000), in this way making it
possible to write consistent definitions, consisting
of a reference to the superordinate concept, genus
proximum, followed by one delimiting characteristic.
In 1993, Gruber defined a framework of
ontological commitments (Gruber, 1993). Later on,
in 1997, the Methontology modeling method was
developed which provided a general guidance to
ontology construction (Lopez et al., 1997). In our
presentation, we will discuss these methods
compared to the methods of terminological concept
modeling and the methods we propose here.
To construct a formal ontology in the domain of
enzyme chemistry, we take as point of departure an
ontology in the biochemical subarea “enzyme
inhibition” created by means of a semi-formal
method (Damhus et al., 2009). The ontology is
intended to be used by subject field specialists for
the purpose of concept clarification. The long-term
goal is to integrate the methodology and the
resulting ontologies and descriptions in the standards
of IUPAC (International Union of Pure and Applied
Chemistry), c.f. (McNaught and Wilkinson, 1997).
This ontology, however, is inconsistent and does not
adhere to the terminological principles that have
been defined by Madsen et al. (2004, 2005).
Therefore we have constructed a new version of
the ontology which we implement in Protégé OWL-
DL. We apply the terminology of terminological
concept modeling when we describe the principles
of terminological ontologies and the OWL
terminology when we describe the OWL
implementation (Horridge et al., 2004). It should be
noted that terminological concept modeling differs
substantially from Object Oriented Modeling (ORM)
that is used for conceptual data modeling.
452
Zambach S. and Nistrup Madsen B. (2009).
APPLYING TERMINOLOGICAL METHODS AND DESCRIPTION LOGIC FOR CREATING AND IMPLEMENTING AN ONTOLOGY ON INHIBITION.
In Proceedings of the International Conference on Knowledge Engineering and Ontology Development, pages 452-455
DOI: 10.5220/0002304904520455
Copyright
c
 SciTePress

Figure 2: The diagram inhibition with subdivision criteria and an artificial layer of concepts.
2 TERMINOLOGICAL
MODELING METHOD
A terminological ontology is a domain-specific
ontology; cf., for example the categorisation of
ontologies by Guarino (1998). We use the term
terminological ontology as a synonym for the term
concept system, which is normally used in
terminology work, e.g. (ISO 704, 2000). The
principles of terminological ontologies are based on
principles that have been used for many years in
terminology work, cf. e.g. Madsen et al. (2004,
2005).
In terminological ontologies, nodes are referred
to as concepts which are defined by means of
concept relations and characteristics that denote
properties of individual referents belonging to the
extension of a concept. In Figure 2 a new version of
the terminological ontology, which was constructed
in the above-mentioned pilot project on ontologies
within the enzyme chemistry domain, is presented.
In terminology work, all kinds of concept relations
are used: type relations (ISA relations), part-whole-
relations and associative relations, such as causal
relations. All relations in Figures 1 and 2 are type
relations. The full ontology also comprises part-
whole-relations and associative relations.
The characteristics of the concepts are presented
as feature specifications in the form of attribute-
value pairs (Carpenter, 1992), e.g. MICHAELIS
CONSTANT: increased. On the basis of these
feature specifications, subdivision criteria are
introduced which provide a good overview and help
the terminologist in writing consistent definitions of
coordinate concepts. Subdivision criteria are in
Figures 1 and 2 represented by means of boxes with
text in capital letters.
Figure 1: Early version of the diagram Inhibition from the
enzyme chemistry project.
According to the terminological principles, two
coordinate concepts must not differ with respect to
more than one characteristic, except if they belong to
a polyhierarchy, where the concepts in question have
two or more superordinate concepts belonging to
different subdivision criteria. In this case the concept
with two or more superordinate concepts is defined
by means of a combination of the characteristics of
the superordinate concepts.
The first version of the ontology did not adhere
to this principle. Figure 1 is a part of this ontology.
In the ontology in Figure 2, we have therefore
introduced a layer of extra concepts: three concepts
that differ with respect to Michaelis constant and
two concepts that differ with respect to Maximum
rate. These concepts are “artificial” and not
important in concept clarification. However, if we
want to adhere to the principle of terminological
ontologies for formalizing the ontology with a view
APPLYING TERMINOLOGICAL METHODS AND DESCRIPTION LOGIC FOR CREATING AND IMPLEMENTING
AN ONTOLOGY ON INHIBITION
453

to consistency checking, this layer of concepts is
important.
3 IMPLEMENTATION IN OWL
The ontology of figure 2 is implemented in OWL-
DL using Protégé 3.4 (Horridge, 2004), c.f. figure 3.
The OWL-file can be found at the website:
ruc.dk/~sz/Inhibition09.owl.
We use OWL DL for its possibility of a fine
grained property structure using e.g. the “hasValue”
operator for datatype properties and the possibility
of more functions in later extensions.
For simplicity, we operate with two kinds of
OWL-properties in order to represent concept
relations, and feature specifications, as mentioned in
section 2.
Type relations and part-whole relations have an
obvious formalization in OWL as ISA relations
among classes and the so called object properties,
respectively.
In addition to these we need to decide which type
of property to use for the implementation of the
feature specifications. In the present implementation,
the features themselves are the data literals “strings
of characters” that are inherited throughout the
ontology. Therefore we have chosen datatype
properties to formalize the feature specifications to
avoid introducing all the values of the feature
specifications as classes.
As an example, see the string “Substrate” in
SubstrateInhibition in figure 3: The class
SubstrateInhibition has the value “Substrate” for the
datatype property: hasInhibitorOfProcess. This
property is inherited through the type relations and
every class has exactly one value for each property.
Figure 3: Conditions for the concept substrate Inhibition in
OWL-DL.
Any feature specification can be represented as a
relation between two concepts, and a concept
relation can be represented as a feature specification.
Therefore we could have considered using object
properties instead, having the possibilities of
creating transitive and symmetric relations. The full
ontology does include such relations, namely part-of
and has-part, which can be transitive. Data
properties are only inherited down in the hierarchy.
The principle of working with only one
delimiting feature specification per concept becomes
feasible in the formal modeling procedure. Siblings
are all separated by characteristics, represented by
feature specifications. This supports a consistent
ontology with a minimum of logical operators for
each predicate since each concept can be described
by its inherited characteristics and one “necessary
and sufficient” description.
This is in line with the suggestion of Minimal
ontological commitment (Gruber, 1993).
4 MODELING PROCEDURE
We suggest that the ontology modeling procedure is
implemented as an iterative process.
We present examples of the steps that were used
for constructing the Inhibition ontology. If this
procedure is followed, the resulting ontology will
have a minimum of necessary and sufficient
conditions. It will consist of defined classes rather
than primitives.
4.1 Terminology Modeling Overview
Below we describe the methods used to construct
formal terminological ontologies.
a. Find sibling concepts related to one
superordinate concept.
b. Identify the characteristics of the concepts.
c. Can the sibling concepts be separated by one
characteristic? If yes, introduce an attribute-
value pair on each concept.
d. Group the siblings by means of one or more
subdivision criteria.
e. If step c-d are not possible and there is a need
for more delimiting characteristics on each
concept, introduce an extra layer of concepts so
that the sibling concepts form part of a
polyhierarchy, i.e. inherit characteristics from
two (or more) superordinate concepts belonging
to two (or more) different subdivision criteria.
f. Define the concepts as classes in e.g. OWL-DL.
g. Define the delimiting features of the sibling
concepts by means of the logical equivalence
operator. If a polyhierarchy is present, the super
classes are added as equivalents.
5 DISCUSSION
The resulting ontology of our modeling procedure as
described in section 4, will, as already mentioned, be
KEOD 2009 - International Conference on Knowledge Engineering and Ontology Development
454

in line with the design criteria related to the
ontological commitments suggested by Gruber
(1993).
Minimal ontological commitment corresponds to
the results of our procedure that lead to the use of
only one operator for each necessary and sufficient
condition.
Clarity is achieved by formulating statements in a
logical axiomatic form. On the other hand, we loose
some readability using Protégé since words and
relations are not formulated in natural language. A
more appropriate and “clear” way of designating
concepts and relations is used in terminological
concept modeling. Also the visualization of
characteristics and subdivision criteria is very clear
and user friendly in terminological ontologies like
the one in figure 2.
Coherence is achieved by using the reasoning
function in Protegé and this application also
facilitates extendability, since other specialists are
able to extend the ontology by using the same
software and the same method to add new concepts.
It may be argued that the encoding in OWL to
some degree suffers from encoding bias. Although
the software generally supports the functionality we
require, a possibility of translating the descriptions
to something more natural language-like would be
appropriate for non-experts.
We propose that the modeling procedure as
described in section 4 should be studied and tested in
development of ontology modeling methodologies
such as Methontology (Lopez, 1997). The
terminological modeling method, described here,
may fine grain the methods of Methontology,
especially in the process of conceptualization,
formalization and implementation.
6 CONCLUSIONS
In this paper, we have presented some central
principles of terminological concept modeling,
applied to an ontology within the subject area of
enzyme chemistry. We have implemented this
ontology in OWL-DL by means of Protégé, and may
conclude that it is possible to implement the basic
features of terminological ontology modeling
(characteristics and concept relations) in OWL-DL,
and in this way it will be possible to check
consistency by using a reasoning function in
Protégé.
Also we may conclude that the visualization
functionality of Protégé does not yet support the
presentation of characteristics and subdivision
criteria in the same way as they are used in
terminological ontologies.
REFERENCES
Baader, F., Calvanese, D., McGuiness, D.L., Nardi, D.
And Patel-Schneider, P.F. editors, 2003. The
Description Logic Handbook. Theory, Implementation
and Applications. Cambridge, UK.: Cambridge
University press.
Carpenter, B., 1992. The Logic of Typed Feature
Structures. Cambridge, Mass.: Cambridge University
Press
Gruber, T.R., 1993. Toward Principles for the Design of
Ontologies Used for Knowledge Sharing. In Formal
Ontology in Conceptual Analysis and Knowledge,
Kluwer Academic Publishers.
Damhus, T., Olesen Larsen, P. Madsen, B.N. and
Zambach, S., 2009. How to work systematically
towards a consistent and codified chemical
terminology – a pilot study. To be published in
Chemistry International, July 2009.
Guarino, N., 1998. Formal Ontology and Information
Systems. In: Formal Ontology in Information Systems,
Proceedings of the First International Conference ,
June 6-8, Trento, Italy, 3-15. Amsterdam: IOS Press.
Horridge, M., Knublauch, H., Rector, A., Stevens, R.,
Wroe, C., 2004. A Practical Guide To Building OWL
Ontologies Using The Protegé-OWL Plugin and CO-
ODE Tools, Edition 1.0, August 27, University of
Manchester, pp. 1-99.
ISO 704, 2000. Terminology work — Principles and
methods. Genève: ISO.
Lopez, M.F., Gomez-Perez, A., Juristo, N., 1997. In
Proc. AAAI Spring Symp. Series, AAAI Press, Menlo
Park, pp. 33-40.
Madsen, B.N., Thomsen, H.E., and Vikner, C., 2004.
Principles of a system for terminological concept
modelling. In: Proceedings of the 4th International
Conference on Language Resources and Evaluation,
Vol. I. Lisbon: 15-18.
Madsen, B.N., Thomsen, H.E., and Vikner, C., 2005.
Multidimensionality in terminological concept
modelling. In: Terminology and Content Development,
TKE 2005, 7th International Conference on
Terminology and Knowledge Engineering,
Copenhagen: 161-173.
McNaught, A.D. and Wilkinson, A., 1997. IUPAC.
Compendium of Chemical Terminology, 2
nd
ed.
Blackwell Scientific Publications, Oxford. XML on-
line corrected version: http://goldbook.iupac.org
(2006-)
Stearns, M.Q., Price, C., K.A. Spackman, and A.Y. Wang,
2001. SNOMED clinical terms: overview of the
development process and project status, in Proc AMIA
Symp, pp. 662–666.
APPLYING TERMINOLOGICAL METHODS AND DESCRIPTION LOGIC FOR CREATING AND IMPLEMENTING
AN ONTOLOGY ON INHIBITION
455