ONTOLOGY-BASED KNOWLEDGE MANAGEMENT
Graphical Query Editor for OWL Ontologies
Markus Schwinn, Norbert Kuhn and Stefan Richter
Institut f¨ur Softwaresysteme in Wirtschaft, Umwelt und Verwaltung, Fachhochschule Trier
Campusallee, 55761 Birkenfeld, Germany
Keywords:
Knowledge management, Ontology, Visual query system, Ambient user support, Graphical query language.
Abstract:
The OnToBau research project aims to provide a way to classify, archive and effectively use business knowl-
edge with the assistance of an ontology-based knowledge archive for small and medium companies from
construction industry. This archive is intended to pro-actively provide users with information to assist them
in their daily business process handling. The system consists of four main parts. The document converters
prepare the different resources (EMails, Paperdocuments, PDFs etc.) that should be stored in the knowledge
archive for the enclosed inference system. The inference system is the core component and extracts the in-
formation from the preprocessed resources. Ontologies provide the necessary domain knowledge. In order
to exploit the available knowledge, a personal agent monitors the current activities of the user and tries to
infer the intention from his behaviors. At certain points it automatically offers the user helpful information.
Again ontologies are used to represent information about the business processes. In addition, the user has the
option to search for information in the archive through the graphical user interface. The importance of simple
query systems has already been identified in the area of database systems. This paper gives an overview of
the OnToBau research project presenting a first approach to visual query for information in the knowledge
archive.
1 INTRODUCTION
The growing importance of computers in the 70s and
the associated opportunity to disseminate information
in digital form, is considered as the turning point to to-
day’s information age. As with any other technolog-
ical revolution in history, the new information tech-
nologies spread within two decades all over the planet
(Castell, 2001). For the first time, it was possible to
produce, to copy and to archive information in a sim-
ple way.
In the early 90s another technological milestone
was passed, that moved the world to a new era of
globalization (B¨oder, 2003). The Internet offered the
opportunity to access and produce information on an
increasing number of websites.
The largest flow of information in the Internet is
produced by e-mails. A study of the Radicati Group
in 2009 stated that in the same year about 247 bil-
lion e-mails per day would be sent, and that this num-
ber would be doubled by the year 2013. However,
approximately 81% of those e-mails can be consid-
ered spam (Lyman and Varian, 2003)(Radicati and
Khmartseva, 2009). This information overload will
be a challenge for many companies in coming years.
This is also covered by the same study: it is estimated
that companies with 1,000 or more employees have to
invest 1.8 million dollars per year in the processing of
e-mails and spam (Radicati and Khmartseva, 2009).
However, not only the flood of information from
the growing e-mail traffic will challenge the com-
panies in the near future. A study of the Gartner
Inc. in 2002 amongst around 300 companies showed
that 96% suffer an information overload. To coun-
teract the information overload, companies will have
to spend 30 billion US-Dollars within the next years
(Goasduff, 2002). A similar investigation of the Ba-
sex Inc. in 2008 stated, that the consequences of in-
formation overload produce costs of about 900 bil-
lion US-Dollars a year for the economy of the United
States, caused by reduced productivity of the employ-
ees in knowledge intensive processes, which have to
spent about 25% of their daily work with searching
for information (Spira, 2008).
Interviews with our project partners from the con-
struction industry provided similar results. In partic-
235
Schwinn M., Kuhn N. and Richter S..
ONTOLOGY-BASED KNOWLEDGE MANAGEMENT - Graphical Query Editor for OWL Ontologies.
DOI: 10.5220/0003491202350240
In Proceedings of the 13th International Conference on Enterprise Information Systems (ICEIS-2011), pages 235-240
ISBN: 978-989-8425-55-3
Copyright
c
2011 SCITEPRESS (Science and Technology Publications, Lda.)
ular, the quotation preparation process has been de-
scribed as a knowledge-intensiveprocess. The reason
is that different resources are used to produce a quo-
tation. Basically, a product database is used. If the
product is not in the database, the employee searches
in product catalogs, previous invoices or web pages.
In addition, business documents are archived in paper
form in many small companies, as document manage-
ment systems are often not tailored to their needs. So
searching for information is often a very time con-
suming task (Schwinn, 2010).
In this paper we describe an architecture, which
should pro-actively provide these small companies
with relevant information stored in their knowledge
archive to assist them in their knowledge intensive
business processes. Because of the enormous variety
of different business processes our system focuses on
processes in the construction domain (e.g., the quo-
tation preparation process). The knowledge base is
built up from different resources a company has to
deal with (e.g. invoices, product catalogs etc.). To
offer the end-user an easy and effective search for in-
formation in the knowledge archive,we present an ap-
proach of a visual query editor.
The remainder of the paper is organized as fol-
lows. First we give an overview of related work in
pro-active knowledge management and visual query
editors in Section 2. The overallsystem architecture is
described in Section 3, with an overview of the main
components. In particular we discuss the visual query
editor in Section 4. Finally, we end up with the con-
clusion and outline some future work.
2 RELATED WORK
In this section we will give an overview of similar re-
search projects considered in process-oriented knowl-
edge management and we present previous work in
the area of graphical query construction, particularly
in the context of ontologies.
2.1 Process-oriented Knowledge
Management
There seems to be no other projects that specifically
consider a knowledge-based process support in con-
struction industry. However, the project DYONIPOS,
which has a strong similarity to the aims of the On-
ToBau research project as described above, tries to
optimize processes in public administration facilities
by providing pro-actively the available knowledge to
the employees (Makolm et al., 2007). DYONIPOS
has adopted a strict process-oriented approach that
moves the focus to the business processes (Tochter-
mann et al., 2006). This approach is appropriate in an
environment with highly structured processes, like in
public administration. Otherwise, in an environment
like the construction industry most of the processes
are semi-structured. In our approach we focus on
the documents and the knowledge contained within
them. Therefore, providing knowledge is more likely
to be tied to the user’s behavior than to rigid processes
(Schwinn, 2010).
2.2 Visual Query Systems
Most approaches to support the end-user with the
query formulation have focused on visual techniques
to hide the target query language like SQL for
databases or SPARQL in the context of ontologies.
(Catarci, 1997) present a classification scheme of 4
different graphical query construction categories of
visual query systems (VQS). The tool described in
this paper belongs to the category of diagram-based
systems, that tend to be the most popular. There have
been a few previous approaches to support a visual
query construction particularly for ontologies. Some
examples include SPARQLViz (Borsje and Embregts,
2006) and NITELIGHT (Russell and Smart, 2008).
SPARQLViz (Borsje and Embregts, 2006) aims to
support the user to query constructions for SPARQL.
The main difference to our approach is the interac-
tion with the user interface. SPARQLViz relies on a
form-based VQS with a wizard-like interface design,
guiding the user through different forms. In contrast,
we present a diagram-based system. There seem to be
no empirical studies on the different VQS categories,
so it is difficult to compare this different approaches.
NITELIGHT (Russell and Smart, 2008) is a VQS
that has much in common with the VQS presented
in this paper and influenced our research to some de-
gree. NITELIGHT supports the user with respect to
the specification of all SPARQL query result forms
(like SELECT, CONSTRUCT etc.). NITELIGHT of-
fers the possibilities of result ordering, filtering and
limiting the results. NITELIGHT is a diagram-based
VQS that offers ontology browsing and drag-and-
drop functionality with a graph-based visualization.
Despite these similarties the following differences do
exist between NITELIGHT and our approach. First,
the VQL presented in this paper is richer compared
to the VQL supported by NITELIGHT. The VQL
presented in this paper offers further possibilities on
property restrictions like range and cardinality restric-
tions (e.g. a person with only invoices before 2010).
Interviews with our project partners from the con-
struction industry revealed that they often search for
ICEIS 2011 - 13th International Conference on Enterprise Information Systems
236
information where partial statements are already
known (e.g. they search invoices with a particular
bathtube or a tender preparation from a specific per-
son). Therefore our VQL supports the query con-
struction including individual statements. The ontol-
ogy browser in NITELIGHT consists of a series of
columns that display the classes and subclasses of the
ontology (Russell and Smart, 2008). Whereas our on-
tology browser provides access to the individuals too.
The current approach is diagram-based, but we are
planning to develop a icon-based system similar to the
result view presented in figure 3. In addition, we use
a live result view to give the user a direct feedback on
his query construction, so he can remark early when
his query goes in the wrong direction.
3 OVERVIEW OF THE
ONTOBAU-ARCHITECTURE
There has been much literature about knowledge
management systems (KMS) within large enterprises
and little information available on KMS within SMEs
(Rasheed, 2005). According to (Rasheed, 2005)
SMEs have special requirements on KMS. Inter-
views with our project partners led to the same re-
sult. The managers in SMEs are in most cases the
owners. The result is, that the decision-making pro-
cess is shorter than in larger companies. They show
a flat and less complex structure, with fewer layers
of management (Wong and Aspinwall, 2004). Pro-
cesses are often not as strongly structured as in larger
enterprises and knowledge is distributed at various
points in the company (file folder, product catalogs,
databases)(Schwinn, 2010). A smaller number of
people within a company is usually united by com-
mon beliefs and values, resulting in shorter and often
less strategic ways for making decisions (Rasheed,
2005). Because of fewer human and financial re-
sources, the introduction of a knowledge management
system should not cause ongoing costs. Especially in
SMEs there are no specialists for knowledge manage-
ment and additional staff costs are not manageable.
The goal of our project is to account for these special
requirements. In summary, we need to consider the
following requirements:
the effort to record analogue documents has to be
minimal and should not disrupt the daily work
intuitive usability
there should be no running costs, in particular no
additional staff cost
knowledge should be extracted from different re-
sources
access to the knowledge archive should be possi-
ble at any time
the extracted knowledge should be linked in an
effective manner to the usual business processes
During the daily work, the employee can decide
whether certain resources should be transferred into
the knowledge base. Those resources (e.g. an e-mail
or a PDF document) are then passed to the OnToBau-
System using the interface of his personal agent or
plug-ins integrated in his office or e-mail software. To
transfer analogue documents to the knowledge base,
we are planning to install special document cameras
directly at the workplace of every employee. In this
way we prevent the scanning process from disturb-
ing the workflows in the company and ensure that
every employee can easily add new resources to the
knowledge base. The purpose of this section is to
give an overview of the architecture of our approach.
As shown in Figure 1, the OnToBau-system consists
of four main parts described in the following subsec-
tions.
3.1 Document Converters
These pre-processing components will prepare the re-
sources for inclusion in the knowledge base. There-
fore the resources are converted to a general represen-
tation language, so called OnToBau Representation
Language (ORL). The pre-processing also includes
various filters (e.g. segmentation filter, part-of-speech
tagger, stop word filter), which are used to simplify
the subsequent processing. The relevant information
is extracted in the inference system component. Inter-
esting e-mails can be convertet into the ORL directly
out of the mail client like Thundebird or Mircosoft
Outlook.
3.2 Inference System
First, a given resource is classified into process related
categories (e.g. an invoice). The domain ontologies
contain information which relevant data have to be
extracted from this categorie. Once the information
has been extracted, it is stored in the knowledge base.
For this purpose, we try to find relevant relationships
between the new data and the existing knowledge in
the knowledge base (e.g. a link between an invoice
and a corresponding quotation). If the resource is an
e-mail with an attachment, it will be removed from
the email and both are integrated into the knowledge
archive.
ONTOLOGY-BASED KNOWLEDGE MANAGEMENT - Graphical Query Editor for OWL Ontologies
237
Figure 1: Overview of the OnToBau architecture.
3.3 Ontologies
The knowledge in the OnToBau-system is represented
by using ontologies. For this, we decided to use the
Web Ontology Language (OWL), which is an W3C
recommendation since February 2004, so we can ben-
efit from existing libraries and tools. The T-Box of
our ontology defines the most relevant concepts and
relations of the construction domain and in general
of business documents, thus providing the OnToBau-
system with the necessary knowledge to decide which
information to extract from the resources (e.g. recip-
ient details, product information etc.). The extracted
information is stored as individuals in the ontology.
3.4 Personal Agent and User Interfaces
The personal agent of the employee performs two
main tasks. First, he should allow the employee to
access the knowledge base by making specific search
queries (e.g. show me all in-voices of Mr. Brown
with an invoice date later than December 2009). For
this purpose we implement a graphical tool for query
construction described in the following chapter. Sec-
ond, the agent should monitor the employees behavior
and provide him pro-actively with information to ful-
fill his task.
4 GRAPHICAL INTERFACE
FOR SEMANTIC QUERY
CONSTRUCTION
According to (Spira, 2008) employees in knowledge
intensive processes have to spent about 25% of their
daily work with searching for information. Particu-
larly in small companies knowledgemanagement sys-
tems are hardly be found due to the reasons mentio-
nend in section 3. To use the corporate knowledge in
the OnToBau-system actively the employees need an
intuitive search facility. Since the knowledge is rep-
resented using OWL ontologies the queries for infor-
mation are restricted to a specialized query language
for ontologies.
Using the example of searching an invoice with
some parameters we will describe our approach: the
knowledge worker is searching for an invoice that was
sent to a recipient who is known as Markus Schwinn
and the invoice date is before january, 1st 2011. He
additionally wants only the invoices that contain or-
dered products of a bathtube or a wash-bowl. In
pseudo code the query string for receiving this infor-
mation from an ontology could look like this:
SELECT ? invoic e WHERE {
? in voi ce a ( in voi ce and
( h asI nvo i ce Rec ipi ent value
MarkusSchwinn ) and
invDate some ? date and
( hasProd uct some
ICEIS 2011 - 13th International Conference on Enterprise Information Systems
238
( bathtub or washbowl ) ) ) .
f i l t e r ( dateTime (? date ) < dateTime
( 20110101T00 :00:00Z ) ) .
}
This seems very familiar to a technically experi-
enced person. But you can not expect that the end-
user is familiar with it too. Visual query systems try
to support an easier way for constructing such a query
string. Like (Russell and Smart, 2008) we decided to
use SPARQL as the target query language for our ap-
proach. The user can select the entities, data and ob-
ject properties of the ontology from a treeview. Con-
trary to NITELIGHT we decided to use individuals
for query constructions too. This reflects the fact that
the knowledge worker often knows some details ex-
actly. In our example he knows the individual of the
invoice recipient. Possibly this will lead to a more
precise and unique query result.
The knowledge worker could drag-and-drop the
objects onto the query editor frame where they are
presented as graphical nodes. They can be dropped
onto the coresponding nodes and are connected au-
tomatically with this node by an edge. The type of
the edges depends on the node types and reflects the
query structure of SPARQL. When needed you can
edit the value of a node to specify cardinalities or
ranges (e.g. when you are searching for a person
with at least three invoices). Figure 2 shows the same
query above but much more intuitive for the knowl-
edge worker.
Figure 2: Example of graphical query structure.
After constructing the graphical query with all the
information he knows, ideally, the knowledge worker
get back one result or a list with all results match-
ing the query. Choosing an item will present the se-
lected individual with all its connections in a well-
arranged graph. The implementation of this resulting
graph is based on the Protege-Plugin Ontograf, which
is slightly enhanced to improve the user experience
(see figure 3). In our example this is the requested
invoice with its connection to the individual ’Markus
Schwinn’. Also two products from this invoice are
noticable, a bathtube and a wash-bowl,represented by
an icon and their name. The result graph shows addi-
tional connections, like the mail which contained the
invoice, the corresponding proposal and the company
sending it.
It is planned but not yet implemented to use this
icon-based approach also for the query editor. In
changing the diagram-based approach to a real graph-
ical approach we undertake a better acceptance on the
user-side because of the re-improved usability.
Figure 3: View of the query result.
A feature called live search is currently under de-
velopment. With every change on the graphical query
frame, like adding or removing nodes or changing the
type of the edges, the knowledge worker should get an
updated list with results matching the current query
term. Thus he get an instant reply if the query makes
sense or not. With this feature the workflow will be
more fluently.
5 CONCLUSIONS AND FUTURE
WORK
Because there are no specific projects supporting the
knowledge management process of companies in the
targeted domain the presented architecture is our way
to provide small and medium companies with an
ontology-based knowledge management system. To
reach this goal we have developed the ORL. This rep-
resentation language is the link between different re-
sources used within companies and the information
extraction. With specific converters the system is able
to transform the resources to a unitary format which
is needed for extraction. Currently we are working
mainly on the extraction step, to enable the system to
ONTOLOGY-BASED KNOWLEDGE MANAGEMENT - Graphical Query Editor for OWL Ontologies
239
build up a knowledge base for a company. This base
will be represented by using ontologies.
An integral part of the OnToBau system which is
not yet im-plemented will be the personal user agent.
The agent will moni-tor the users activities and pro-
vide him with the relevant infor-mation for the pro-
cess in real-time. To achieve this, it needs underlying
behavior patterns and must try to anticipate the users
intention.
The editor for the visual query construction is still
under development. Currently it has a diagram-based
approach, but it will change to an icon-based visual
style. This might improve the user experience with
our tool because there will be no ”visual gap” be-
tween query construction and result graph. We have
not performed any user evaluations with our project
partners, but aim to undertake such studies in the near
future. Another extension is the live search function-
ality. When finished it will provide the knowledge
worker with a direct response on what he does in the
query editor. Currently the query editor and the pre-
sentation of the result is divided in two program parts.
For a fluently workflow it will be neccessary to com-
bine this in one part.
Our research project focuses on the needs of con-
struction industry but of course this approach could
be adapted in other domains dealing with ontologies.
ACKNOWLEDGEMENTS
This work is supported by the German Ministry of
Research and Technology under grant FKZ 1731X09.
REFERENCES
B¨oder, J. (2003). Das Internet und seine Folgen - Was
Luthers Reformation und unsere Zukunft gemeinsam
haben. http://www.intelligenzia.de/Texte/Internet.pdf.
17. August 2009.
Borsje, J. and Embregts, H. (2006). Graphical Query Com-
position and Natural Language Processing in an RDF
Visualization Interface. Bachelor, Erasmus University
Rotterdam.
Castell, M. (2001). Das Informationszeitalter. Wirtschaft,
Gesellschaft, Kultur, volume 1. Leske + Budrich Ver-
lag.
Catarci, T. (1997). Visual Query Systems for Databases: A
Survey. Journal of Visual Languages & Computing,
8(2):215–260.
Goasduff, L. (2002). Gartner Says 90 Percent of
Businesses Suffer from Information Overload.
http://www.gartner.com/5 about/press releases/2002 -
05/pr20020507c.jsp. 24. August 2009.
Lyman and Varian (2003). How Much Information?
http://www2.sims.berkeley.edu/research/projects/how-
much-info-2003/printable report.pdf. 19. August
2009.
Makolm, J., Weiß, S., and Reisinger, D. (2007). Proactive
knowledge management: the DYONIPOS research
and use-case project. In Janowski, T. and Pardo, T. A.,
editors, Proceedings of the 1st international confer-
ence on Theory and practice of electronic governance
(ICEGOV ’07), pages 85–88, New York, NY, USA.
ACM.
Radicati, S. and Khmartseva, M. (2009). Email Statistics
Report, 2009-2013. http://www.radicati.com/wp/wp-
content/uploads/2009/05/email-stats-report-exec-
summary.pdf. 19. August 2009.
Rasheed, N. (2005). The Impact Of Knowledge Manage-
ment On Smes. Technical report, Knowledgeboard.
Russell, A. and Smart, P. R. (2008). NITELIGHT: A Graph-
ical Editor for SPARQL Queries. In International Se-
mantic Web Conference (Posters & Demos).
Schwinn, M. (2010). Automatic Information Extraction
from Documents for Setting Up a Knowledge Archive
as Basis for an Automated Tender Preparation. Mas-
ters thesis, University of Applied Sciences, Trier.
Spira, J. B. (2008). Information Overload: Now $ 900
Billion - What is Your Organizations Exposure?
http://www.basexblog.com/2008/12/19/information-
overload-now-900-billion-what-is-your-
organizations. 10. December 2010.
Tochtermann, K., Reisinger, D., Granitzer, M., and Lind-
staedt, S. (2006). Integrating Ad Hoc Processes and
Standard Processes in Public Administrations. In Pro-
ceedings of the OCG eGovernment Conference, Linz
(Austria),.
Wong, K. and Aspinwall, E. (2004). Characterizing knowl-
edge management in the small business environment.
Journal of Knowledge Management, 8:44–61.
ICEIS 2011 - 13th International Conference on Enterprise Information Systems
240