ONTOLOGY MODELING TOOL USING CONCEPT
DICTIONARY AND INFERENCE
Yoichi Hiramatsu
Galaxy Express Corporation
1-18-16 Hamamatsucho, Minatoku,
Tokyo 105-0013, Japan
Keywords: Ontology Modeling, Editing Tool, Concept Dictionary, EDR, Lexical Dictionary, WordNet, Inference
Algorithm, Web service, Enterprise System, Common Lisp, I-S-A semantic link, Operation Support System
Abstract: The usefulness of ontology is strongly dependent on a knowledge representation policy and its maintenance.
The subject of knowledge representation and modeling tools has been one of the most exciting themes
among ontology scientists. Some ontology editing tools originated and developed in the field of expert
systems, and others were designed originally by ontology research groups. Key features of the newly
implemented tool are: (a) reference to a concept dictionary (EDR, WordNet) to ascertain word semantics,
and (b) use of an inference algorithm (MOP) provided by Schank et al. Satisfactory results were obtained in
the application of ontology modeled by the present tool. We will discuss how our tool was constructed and
describe applications using the tool to achieve solutions for enterprise integration. This work was developed
as a part of the project entitled “Operation-support system for large-scale system using information
technology” (Koide et al., 2003) for the Japanese Government IT Program, period 2002-2005.
1 INTRODUCTION
One of the design goals of the ontology modeling
tool is to provide an interactive and graphical facility
for constructing ontology representation files. A
graphical user interface (GUI) should work as an
advisory system wizard providing developer
guidance during the construction of class hierarchies
and relationships. This paper deals with one aspect
of these goals: that is, implementation of a modeling
tool (named “MOP Editor”) connected to a concept
dictionary and to an inference algorithm. Despite the
existence of attractive tools to model ontologies, we
have persisted in developing the necessary tool for
our specific requirements.
The category of modeling tools described herein
sh
ould gain considerable importance for the
concretization of novel software architecture such as,
for example, those which model and generate
business-to-business (B2B) applications supporting
decentralized and dynamic electronic agents (Alloui
et al., 2003).
If we find a way to access an external source of
referen
ce to determine ontology, we can save time
during the editing phase. We have decided to use an
online concept dictionary (EDR, 2003) and a lexical
dictionary (WordNet, 2003) to refer to for concepts
of Japanese and English terms. And, if we devise a
way to interactively test functionality of the
“in-construction” ontology during the construction
phase, the ontology quality is improved significantly
in comparison with that obtained without interactive
tests. Thus, we have chosen an inference engine
known as MOP (Memory Organization Package)
algorithm introduced by Schank et al. (Riesbeck et
al., 1989; Schank et al., 1994) in order to execute
ontology evaluation tests. MOP is a kind of
frame-based knowledge formalism that holds class
hierarchy, in which concepts and instances are
represented by mop objects.
The effectiveness of MOP Editor and the
con
structed ontology were evaluated against
objective design criteria using three kinds of Web
services: message generation, synonym retrieval,
and ontology sharing. Web services are distributed
around the system and work collaboratively.
2 ONTOLOGY AND DOMAIN
FEATURES
Ontology is a set of conceptual building blocks of
225
Hiramatsu Y. (2004).
ONTOLOGY MODELING TOOL USING CONCEPT DICTIONARY AND INFERENCE.
In Proceedings of the Sixth International Conference on Enterprise Information Systems, pages 225-230
DOI: 10.5220/0002603002250230
Copyright
c
SciTePress
knowledge in a determined application area. It
provides a domain of discourse for knowledge
sharing among computers. Ontology enables a
number of machines to share their knowledge: for
example, for information retrieval systems.
Concretely, ontology enumerates concepts, attributes
of concepts, relationships among concepts, and
constraints on these relationships, thereby defining
the knowledge reference structure of the considered
domain.
Our domain is ground support equipments for
rocket launching at a rocket launch facility at
Tanegashima Space Center in Japan. Entry terms of
the domain are the names of the following objects:
fluid-pump, air-compressor, pipe, junction,
storage-device, electrical-cable, fuel-tank,
oxygen-tank, valve, actuator, fuel, oxygen, vaporizer,
sensor, pressure-controller, etc.
Abstract concepts that correspond to the entry
terms are arranged in such a way as to compose a
class hierarchy in which superclass and class are
connected by an I-S-A semantic link. For example,
“pressure-controller is-a-kind-of controller”.
Ontology can be accurately represented by the class
hierarchy of abstract concepts plus actual instances.
Instances are tagged names of real objects, such as
sensor-AA, controller-101, controller-102, and so on.
Instances are attached to ending points of the leaf of
the class hierarchy. In general, the number of
instances is estimated to be one order greater than
the number of classes. In our present work, we have
selected a quantity of over a hundred for classes and
over a thousand for instances.
3 CURRENT ONTOLOGY
MODELING TECNIQUES
Free software tools and commercial products for
ontology modeling have evolved during the last few
years. Some examples are Ontolingua, Protégé2000,
OntoEdit and Hozo. These tools accept entry terms
and create an I-S-A semantic link for classes,
generating text-type files of class hierarchy. Most of
them are equipped with plug-in features for
importing other ontologies written in XML, RDF,
DAML, OWL, etc.
Ontolingua (Farquhar et al., 1996; Ontolingua,
2003) is a set of tools and services that supports the
process of achieving consensus on common shared
ontologies by geographically distributed groups.
Ontolingua makes use of the world-wide web to
enable wide access and provide developers with the
ability to publish, browse, create, and edit ontologies
stored on an ontology server.
On the other hand, the Protégé system (Gennari
et al., 2002; Noy et al., 2001; Noy and Musen, 2003)
is an environment for building knowledge-based
systems, mainly in the field of domain ontology.
Protégé runs on a variety of platforms, supports
customized user interface extensions, incorporates
the Open Knowledge Base Connectivity (OKBC)
knowledge model, and interacts with storage formats
such as relational database, XML and RDF.
Similarly, OntoEdit (Sure et al., 2002; OntoEdit,
2003; Maier et al., 2003) is another ontology editing
environment that supports the development and
maintenance of ontologies by graphical means.
OntoEdit is built on top of an internal ontology
model, enabling therefore as much neutral modeling
as possible for concepts, relations and constraints. It
is equipped with a GUI menu entry (in which
developers can choose namespace) and a back-end
inference engine.
Another interesting tool that we have
investigated is Hozo (Kozaki et al., 2002; Kozaki et
al., 2000). This has different features than
Ontolingua, Protégé or OntoEdit. Hozo is based on
an ontology theory of role-concept in which the tool
can distinguish concepts dependent on particular
contexts from the so-called basic concept, and can
manage the correspondence between a wholeness
concept and a relatioship concept.
The newly implemented MOP Editor differs
from the modeling tools described above in the
following points: (a) it is coded in Common Lisp
(CL) and Common Lisp Object System (CLOS),
thereby allowing capability for dynamic
maintenance, (b) it makes reference to the concept
dictionary to help developers to build as general an
ontological model as possible, and (c) it supports the
inference engine that will be used to check the
appropriateness of the built ontology.
4 IMPLEMENTED MODELING
TOOL
4.1 Configuration of Tool
Figure 1 is a process view of the MOP Editor. The
GUI facility was implemented using the Integrated
Development Environment (IDE-CG) of the CL.
The MOP algorithm was developed using a set of
functions of CLOS (Koide and Kitamura, 2002).
Although there are a lot of respectable programming
languages available, we think that CL is, in a sense,
the mother of all languages and is highly efficient.
The main reason for this statement is that CL
includes a complete theory of computation by
treating code and data within a single and uniform
system.
ICEIS 2004 - INFORMATION SYSTEMS ANALYSIS AND SPECIFICATION
226
Link characterization among concepts is based
on the I-S-A (is-a-kind-of) semantic link, as adopted
in EDR and WordNet.
4.2 Class Hierarchy
The class hierarchy mentioned here is a
classification hierarchy. Concepts are connected by
an is-a-kind-of relationship. This relationship can be
seen as the following: if X is a kind of Y, then X is a
specialization of Y, while Y is a generalization of X.
Therefore, our ontological representation is based on
the classification hierarchy.
4.3 Data Organization
MOP Editor organizes the developers data in a form
of project unit (file named *.mprj). Each project
contains all the information that developers have
typed in, that is; the ontology file (named *.mont)
and the instances file (named *.mins). The ontology
file and instance file are composed of Lisp
S-expressions for mop objects and mop instances,
respectively. The following are examples of
ontology files and instance files:
<ontology file>
(defmop mop-object (superclass superclass …)
(role1 filler1)
(role2 filler2)
(role3 filler3)
(role4 filler4)
…)
<instance file>
(definstance mop-instance (superclass)
(has-a mop)
(part-of mop)
…)
The schematic shown in Figure 2 clarifies the
input and output files of MOP Editor. Projects file
*.mprj can be loaded dynamically by developers
during ontology maintenance services, without any
necessity for system shutdown.
Fig
4.4 Utilization of Concept Dictionary
MOP Editor utilizes two kinds of online dictionaries
as a reference source to build the top, domain and
task ontologies. The flowchart in Figure 3 shows
how developers use the parser software and
dictionaries to add new concepts. Developers can
switch between the two dictionaries according to
their needs when looking for new concepts.
EDR
EDR is a machine tractable dictionary of
Japanese words used in the research field of natural
language processing. The words resource is
organized in the form of data records. The main
feature of EDR is that its sub-dictionaries are
interconnected via concepts using concept ID tags.
In this work, we have used three sub-dictionaries
of EDR: (a) a Japanese Word Dictionary, (b) a
Concept Classification Dictionary, and (c) a
Headconcept Dictionary. The role of the Japanese
Word Dictionary is to describe the correspondence
ure 1 : Process View of MOP Editor
Web Service
(server)
Ontology Server
(front-end)
Web Service
(client)
Socket
Web Service
MOP
Editor
E D R
WordNet
(defmop
…)
(definstance
…)
Ontology Server
(back-end)
Developer
Saving data
Loading data
reference
Figure 2 : Schematic of Input / Output
entry.txt
*.mprj
(*.mont , *.mins)
Input file
Output file
loop applied for existent projects
*.mprj
(*.mont , *.mins)
MOP
Editor
Figure 1: Process view of MOP Editor
Figure 2: Schematic of Input/Output
Figure 3 : Flowchart
Technical
Documentation
Parser
Stop
Class hierarchy
Start
Edit slots
*.mprj
*.mont
*.mins
“class” &
“instance”
Dictionary of Concept:
Text file
reference
reference
E D R
WordNet
Figure 3: Flowchart
ONTOLOGY MODELING TOOL USING CONCEPT DICTIONARY AND INFERENCE
227
between Japanese words and the concepts
represented by these words, and also to provide
grammatical information for the word when used
with a given meaning. Each record is composed of
the following fields: record number, headword,
invariable portion of the headword and pair of
adjacent attributes, Kana notation, pronunciation,
syntactic category (noun, verb and particle),
syntactic tree, conjugational information, surface
case information, aspect information, word function
information, concept ID, English headconcept,
Japanese headconcept, English concept explanation,
Japanese concept explanation, usage, frequency, and
management information.
Thus, the purpose of the Concept Classification
Dictionary and the Headconcept Dictionary is to
provide concepts for the Japanese Word Dictionary.
The first contains a classification of concepts having
a super-sub relation, and the data record is composed
of the following fields: record number, concept ID of
the superconcept, concept ID of the subconcept, and
management information. The second one gives a
description of each concept in words, and the record
is composed of the following fields: record number,
concept ID, English headconcept, Japanese
headconcept, English concept explanation, Japanese
concept explanation, and management information.
Both dictionaries contain about 400,000 concepts.
MOP Editor uses a concept ID to connect
concepts among these sub-dictionaries. The ontology
developer would display the superclass and class of
concepts when he is looking for a relationship that
can better represent the subsumption relationship (or
I-S-A semantic link) in the treated domain.
WordNet
WordNet is an online lexical resource of English
language used by linguistic scientists. WordNet
consists of lexicographer files that organize nouns,
verbs, adjectives and adverbs into groups of
synonyms called synsets (synonym sets) and
relationships between this synset and other synsets.
Nouns and verbs are organized into a hierarchy
based on the hypernymy/hyponymy and
holonymy/meronymy relationships between synsets.
Here, hypernymy and hyponymy mean, respectively,
superordinate and subordinate relations. For
example, Y is a hypernym of X, if X is a (kind of) Y.
Holonym is the name of the whole of which the
meronym names a part. For example, Y is a holonym
of X, if X is a part of Y. WordNet shows how each
word is linked to others, as for example when the
developer types in the word valve, he gets not only
the synonyms definition, but the hypernyms (a valve
is a kind of what?), meronyms (what are the parts of
a valve?), and more.
4.5 Behavior of Inference Engine
MOP is a frame-based formalism based on the
Schank’s theory about how human memory is
organized, and attempts to remind computers the
way people are reminded.
In MOP algorithm, we use a global symbol named
mop to represent a CLOS object. The mop object can
have multiple superclasses and a set of attributes
represented by slots, such as (role1 filler1), (role2
filler2), (role3 filler3), ... The role is a slot-name
and the filler is a slot-value. The filler takes values
of the type string, integer, double-float and own mop
object. The class hierarchy is a representation of
abstract concepts using mop objects. When
developers create a new mop object, MOP Editor
evaluates this mop object in order to avoid eventual
definition errors of the mop object. MOP Editor still
checks whether the attributes represented by the pair
of role and filler are consistent inside the ontology
file and instances file.
A new mop object is created if all the fillers have
been approved through this error checking
evaluation test. Note that if the filler already exists,
then the previous defined filler is used for the
evaluation test. On the other hand, if the filler does
not exist, then the evaluation test uses the filler of
the present mop object. Therefore, MOP algorithm
differs from the frame-based system in the following
points: (a) the newly created filler behaves as a mop
object, (b) the filler is also structured hierarchically,
(c) mop hierarchy is flexible, i.e., the knowledge
representation becomes more and more detailed
every time new mop objects are added to the class
hierarchy. MOP uses slots to place the new instance
at an appropriate location inside the class hierarchy.
Figure 4 shows a view of the implemented MOP
Editor.
Figure 4 : A View of MOP Editor
Figure 4: A view of MOP Editor
ICEIS 2004 - INFORMATION SYSTEMS ANALYSIS AND SPECIFICATION
228
5 EVALUATION OF TOOL AND
ONTOLOGY
Functionality of MOP Editor and appropriateness of
the constructed ontology were evaluated using three
kinds of Web services: message generation service,
synonym retrieval service, and ontology sharing
service. Web service is a software system designed
to support interoperable machine-to-machine
interaction in a distributed and collaborative
environment, equipped with an interface described
in a machine-processable format under the SOAP
protocol.
5.1 Message Generation Service
Message generation is one of the simplest Web
services. The Web client of a message generation
service queries for generation of messages
comprehensive to humans by sending a set of
keywords in direction to the ontology server. The
ontology server uses an adequate template and sends
a reply message to the client. The role of the
ontology is to provide a reference base of
information retrieval to compose grammatically
correct messages that better fit the client queries.
For example, for a given set of keywords such
as: part = controller, device = LNG drain control
valve and event = abnormal closing, the generated
message would be “Because of malfunction of the
controller at LNG drain control valve, abnormal
closing was detected”.
This message is sent back to the Web client. As
shown in Figure 5a and Figure 5b, we confirmed
that our ontology is capable of generating messages
comprehensive to humans for several contextual
situations in the range of computer simulation.
(defmop Context (Root) ())
(defmop Cause-Candidate (Context) ())
(defmop Device-Part-Abnormal-Cause-Candidate (Cause-Candidate)
((device
Device)
(part
Part)
(event
Abnormal-Event)
(phrase ‘(Because of the malfunction of the (part) at
(device), (event) is possible.))))
Generated Message
“Because of the malfunction of the controller at
LNG drain control valve, abnormal closing is possible.
(slots->mop ‘((device
5.2 Synonym Retrieval Service
Synonym retrieval is a Web service for information
retrieval. The Web client queries for a synonym that
has the same or similar meaning as a given word; for
example, for liquid-oxygen the synonym would be
oxidant, for temperature-sensor the synonym is
thermocouple or thermometer, and so on.
Furthermore, in specialized fields engineers use
different technical terms to indicate identical or
equivalent physical objects, such as
fuel-feeding-system and launching-tower, or even to
differentiate one particular physical hazard from
another more general one, such as fatigue-crack and
crack. Since operational tasks for rocket launching
comprise a large-scale system embracing hundred of
engineers of different divisions and companies, a
Web-searching service for adequate synonyms gains
significant importance during diagnostics.
5.3 Ontology Sharing Service
Use of a fixed source of reference such as EDR or
WordNet makes the built ontology transferable to
other database servers distributed on the network.
Thus, sharing of ontology becomes possible because
I-S-A link and hypernymy/hyponymy or
holonymy/meronymy relationships can be
considered “invariable” during a long period of time.
The Web client of an ontology sharing service
queries for a copy of one part or a whole ontology
stored in the ontology server. A copy of class
hierarchy will be shared with other data servers. In
this work, we have tried a sharing service between
an ontology server and a so-called large-scale
multimedia filing data server. A sharing service is
used for a rapid keyword-based search of
multimedia data (documentation, drawings, logs,
sound, pictures and movies) during operational tasks
and diagnostics at the rocket launch facility.
FCV-5201)
Controller)
Close-Abnormal))
(part
(event
‘(Context) nil)
This input automatically
creates an instance of
Device-Part-Abnormal-
Cause-Candidate.
Then, generate message…
Figure 5 b : Message Generation
Input Data
Cause-Candidate
(defmop Device-Role (Root) ())
(defmop Part (Device-Role) ())
(defmop (Part)
((phrase ‘(controller))))
(defmop Event (Root) ())
(defmop Abnormal-Event (Event) ())
(defmop (Abnormal-Event)
((phrase ‘(abnormal closing))))
(defmop Object (Root) ())
(defmop Device (Object) ())
(defmop Valve (Device) ())
(defmop Flow-Control-Valve (Valve) ())
(definstance (Flow-Control-Valve)
((phrase ‘(LNG drain control valve))))
Fi n
context
device
part
event
Controller
Close-Abnormal
FCV-5201
Controller
Close-Abnormal
FCV-5201
Figure 5b: Message Generation
gure 5 a : Message Generatio
Figure 5a: Message Generation
ONTOLOGY MODELING TOOL USING CONCEPT DICTIONARY AND INFERENCE
229
6 SUMMARY AND FUTURE
WORK
MOP Editor demonstrated ability in constructing
ontology applicable to Web services for message
generation, synonym retrieval, and ontology sharing.
We have confirmed throughout our computer
simulation that the tool can be used for graphically
modeling class hierarchy and semantic links.
Furthermore, we have assured that ontological
models provided by EDR and WordNet work
relatively well for the presented Web services.
Future work is needed to enhance GUI facility
and MOP functionality to incorporate coming
themes such as text extraction from any sources, task
ontology, ontology distinguishing, integration of
different ontologies, and plug-in for database
connectivity.
ACKNOWLEDGEMENT
This work has been supported in part by the Ministry
of Education, Culture, Sports and Technology
(MEXT) of Japan. The authors are grateful to
Professor Dr. R. Mizoguchi and Research Associate
Dr. K. Kozaki of Osaka University for providing
helpful comments on ontology methodology. Many
thanks to the technical group of Franz Incorporated
for valuable advice on Allegro’s IDE-CG. Many
thanks also to all the colleagues of Galaxy Express
Corporation for their enthusiasm and contributions
in developing the project.
REFERENCES
Alloui, I., Megzari, K., Oquendo, F., 2003. Modeling and
generating business-to-business applications using an
architecture description language-based approach. In
ICEIS-2003, 5
th
International Conference on
Enterprise Information Systems. ICEIS Press.
EDR (2003)
http://www.jsa.co.jp/EDR/
Communications Research Laboratory. Web site.
Farquhar, A., Fikes, R., Rice, J., 1996. The Ontolingua
Server: A Tool for Collaborative Ontology
Construction. Knowledge Systems Laboratory,
Stanford University.
Gennari, J. H., Musen, M. A., Fergerson, R. W., Grosso,
W. E., Crubezy, M., Eriksson, H., Noy, N. F., Tu, S. W.,
2002. The evolution of Protégé: an environment for
knowledge-based systems development. In
International Journal of Human-Computer
Interaction.
Koide, S., Kitamura, Y., 2002. MOP3: Memory
Organization Package on Meta Object Protocol. In
ILC-2002, International Lisp Conference, San
Francisco, USA.
Koide, S., Nishio, H., Kitamura, Y., Gofuku, A.,
Mizoguchi, R., 2003. Operation-support system for
large-scale system using information technology. In
ICEIS-2003, 5
th
International Conference on
Enterprise Information Systems. ICEIS Press.
Kozaki, K., Kitamura, Y., Ikeda, M., Mizoguchi, R., 2000.
Development of an environment for building ontologies
which is based on a fundamental consideration of
“relationship” and “role”. In PKAW2000, Proceedings
of the Sixth Pacific Knowledge Acquisition Workshop.
Kozaki, K., Kitamura, Y., Ikeda, M., Mizoguchi, R., 2002.
Hozo: an environment for building/using ontologies based
on a fundamental consideration of role and
relationship. In EKAW2002, 13
th
International
Conference Knowledge Engineering and Knowledge
Management.
Maier, A., Schnurr, H., Sure, Y., 2003. Ontology-Based
Information Integration in the Automotive Industry. In
ISWC-2003, 2
nd
International Semantic Web
Conference.
Noy, N. F. and McGuinness, D. L., 2001. Ontology
Development 101: a guide to creating your first ontology.
In SMI technical report (SMI-2001-0880), Stanford
University.
Noy, N. F. and Musen, M. A., 2003. Ontology versioning
as an element of an ontology-management framework. In
IEEE Intelligent Systems.
Riesbeck , C. K. , Schank, R. C., 1989. Inside case-based
reasoning. Lawrence Erlbaum Associates, Publishers.
New Jersey.
Schank, R. C., Kass, A., Riesbeck, C. K., 1994. Inside
case-based explanation. Lawrence Erlbaum Associates,
Publishers. New Jersey.
Sure, Y., Erdmann, M., Angele, J., Staab, S., Studer, R.,
Wenke, D., 2002. OntoEdit: collaborative ontology
engineering for the semantic Web. In ISWC-2002, 1
st
International Semantic Web Conference.
OntoEdit (2003)
http://www.ontoprise.de
Ontoprise GmbH. Web site.
Ontolingua (2003)
http://www-ksl-svc.stanford.edu/
http://ontolingua.stanford.edu/
Knowledge Systems Laboratory, Stanford University.
Web site.
WordNet (2003)
http://www.cogsci.princeton.edu/~wn/
ftp://ftp.cogsci.princeton.edu
Cognitive Science Laboratory at Princeton University.
Web site.
ICEIS 2004 - INFORMATION SYSTEMS ANALYSIS AND SPECIFICATION
230