COLLABORATIVE ONTOLOGIES AND ITS VISUALISATION IN
CSCW SYSTEMS
Michael Vonrueden
University of Paderborn, Faculty of Electrical Engineering, Mathematics and Computer Science
Fuerstenallee 11, 33102 Paderborn, Germany
Thorsten Hampel
University of Paderborn, Faculty of Electrical Engineering, Mathematics and Computer Science
Fuerstenallee 11, 33102 Paderborn, Germany
Keywords:
Collaborative Ontology Creation, Knowledge Representation, Userinterface, Semantic Web.
Abstract:
The goal of semantic structures and especially the semantic web is to simplify knowledge retrieval in computer
based systems. The Visual Cooperative Ontology Environment - short visCOntE - aims to support the process
of a collaborative ontology creation and the mapping of an individual’s mental map into a computer based
knowledge representation. Due to the collaborative and graphical approach many requirements have to be
considered to establish such a project. Beneath a deep description of visCOntE and possible usage scenarios,
the question what is needed for a successful collaborative ontology creation and which functions a system
should make available will be determined in detail.
1 INTRODUCTION
The problem of information overload is present in
nearly every digital based system. The difficulty to
gain an overview and to separate the wanted from
the unwanted information, or more precisely, to trans-
form it into personal knowledge is often an enormous
cognitive performance. The latest ambitions of the
W3C towards the semantic web shows a new way to
handle these tasks (W3C, 2004a). The use of ontolo-
gies can make it easier to find, process and associate
informations. But as an ontology is similar to a men-
tal model of an individual, a monolithic mental model
will show only a unique view to the world, which
will, in some cases, be not comprehensibly for oth-
ers. Therefore the creation of ontologies should cover
the involvement of different parties to participate and
construct an ontology in a collaborate process.
This paper outlines the problems with ontologies
which are developed in a one to many relation and
tries to find new ways for a collaborative creation of
ontologies. After an analysis of what is needed for a
collaborative ontology creation, the current develop-
ment of a tool called visCOntE will be described. vis-
COntE (Visual Cooperative Ontology Environment)
- part of the
open
sTeam (Structuring Information in
Teams) open source environment, developed at the
University of Paderborn - is designed to edit and
browse ontologies in a collaborate environment.
The underlying high flexible
open
sTeam server ar-
chitecture, provides manifold approaches to create
and structure knowledge in a cooperative manner.
Users can share and annotate their knowledge (e.g.
Documents) or separate resp. order it by creating spe-
cial rooms and access rights.
open
sTeam’s communi-
cation facilities like the shared whiteboard supports
the synchronous building of a cooperative informa-
tion space (Hampel and Keil-Slawik, 2002). The vis-
COntE project is the very first step to integrate a new
approach of ontology based knowledge management
into this environment.
2 SEMANTIC STRUCTURES
A human sees the world not as a line up of single ob-
jects but rather as an interconnected ensemble. The
ability to structure natural objects and artifacts in hi-
erarchies and classes is one part of the preceding step
in a construction of a semantic map. The other highly
cognitive performance is to relate the items of a tax-
onomy to other items and to reason their associations.
To achieve this relations a kind of logical reasoning
has to be performed. With the resulting semantic
structure new semantical integrations can be adopted.
But the question is how to integrate such complex
structures into a computer based system. The struc-
ture of an ontology allows the creation of special
knowledge domains, which split the complex type of
a human mental model into specialized parts, which
294
Ronaghi F. (2005).
INTEGRATED PERFORMANCE MANAGEMENT.
In Proceedings of the Seventh International Conference on Enterprise Information Systems, pages 294-299
DOI: 10.5220/0002551402940299
Copyright
c
SciTePress
have a common base. These ontologies cover (instead
of a whole representation ) only a fine granulated part
of it. This makes the processing of knowledge easier
and provides more reusability for distributed ontolo-
gies.
3 COLLABORATION AND
ONTOLOGIES
”The Semantic Web is an extension of the
current web in which information is given well-
defined meaning, better enabling computers and
people to work in cooperation. (Tim Berners-
Lee, 2001)
As mentioned above, ontologies are similar to an in-
dividuals mental model of a special domain and/or a
view of the world. The process to create such a men-
tal model is far more than a one-time procedure or a
simple adaption of other individual’s models. More-
over the development of a mental model is always in a
state of continuous movement associated with social
aspects like discussions, correction of wrong assess-
ments or simply: questions and answers. The rule of
a mental models creation could be described as fol-
lowed: Starting from a generic point of knowledge the
new adopted knowledge will become more and more
special, but never reaches a final state.
This let us assume, that for a successful building
of knowledge two parts are essential: continuousness
and endlessness. The fact is, that this parts are not
considered in ontologies which are created from one
or a few authors for many users or recipients and those
which claim a final state. Because of the non existing
involvement of the future users and the missing social
aspects, a collaborative approach is needed which will
support the evolution of an ontology similar to a men-
tal model.
A collaborative created ontology could bear ad-
vantages with regard to the acceptance and compre-
hension of a common knowledge domain by all par-
ties. The ability to create, modify and contribute an
individual knowledge could further affect the atten-
dance of the whole process and the gain of new or
renewed knowledge which come along with discus-
sions. Moreover a hidden advantage exists in the in-
terpretation of current states of an ontology, that could
state out the actual knowledge of a community and
outlines the way of creation up to an actual snapshot.
The resulting questions is, what should be consid-
ered to transform the idea of a collaborate creation of
ontologies into a real word scenario. In relation to the
previous conclusions, the following keywords should
be reasonable represented in this scenario: visualiza-
tion, comprehension, social exchange and usability.
4 visCOntE - VISUAL
COOPERATIVE ONTOLOGY
ENVIRONMENT
visCOntE - part of the
open
sTeam (Structuring Infor-
mation in Teams) environment, developed at the Uni-
versity of Paderborn (Hampel and Keil-Slawik, 2002)
- aims to be a web-based graphical front-end for the
collaborative creation and modification of ontologies.
Intended are three possible kinds of editable ontolo-
gies. Due to the capabilities of the
open
sTeam Server
it is possible to provide a very global ontology for the
whole (server-wide) community, the creation of spe-
cial rooms and groups permit a shared ontology for
related users and the personal user space for private
one. In general visCOntE should act as the default
application for files in the common W3C Ontologie-
Format OWL (W3C, 2004b) , which makes it very
flexible in importing existing ontologies. The envi-
ronment itself is splitted into a client and a server
module. The client, implemented in Scalable Vec-
tor Graphics, combines an ontology browser with di-
rectly accessible edit-components. With this browser
it is possible to structure the taxonomy of the ontol-
ogy and to add, remove and edit elements such as
classes and individuals or their interconnection. The
server module, responsible for storing and retrieving
the ontology data in OWL Format, communicates to
the client over a simplified XML-Format to avoid time
consuming processes. The following outlines the in
section Collaboration and Ontologies mentioned key-
words in relation to the developed tool.
4.1 Server Module
Like the
open
sTeam server itself, the server-module
is written entirely in the Pike Scripting Language.
Since Version 7.6, the Pike API offers basic capa-
bilities for the handling of RDF (Ora Lassila, 1999)
and OWL structures (Michael K. Smith, 2004) . The
visCOntE server-module builds up on this basic func-
tionality by adding a more sophisticated and linked
level which represents the structure of the low level
OWL-Elements like Classes, Individuals and Proper-
ties. This linked structure extends the meta informa-
tions of these elements, so that a class for example
knows which individuals are derived from itself or
an individual can validate if it belongs to the inher-
itance line of a special class. A controllstructure acts
as an intermediate between the elements and the tex-
tual representation on the one hand and the connec-
tion to the Pike API on the other. Further it handles
the ontology access and versioning over platform spe-
cific storage handlers.
Due to the extended element representation the ar-
chitecture supports the use of actions which can be
COLLABORATIVE ONTOLOGIES AND ITS VISUALISATION IN CSCW SYSTEMS
295
performed on hierarchical and ontology specific data
structures. These actions will take place only on the
represented elements and not on the ontology itself,
so moving a class to a new position in the tree, adding
an individual of a class to the ontology or getting an
individual with a specific property are possible with-
out touching the existing data. The above mentioned
ontology control object will save the changes only on
a explicit save call. The execution of a single ac-
tion produces a result object which is provided either
as a simplified hierarchical XML Format or in a raw
format of the above mentioned OWL-Elements. The
action model can be extended by customized actions
which makes it easier for custom interfaces to access
and edit an ontology in a very special way and with
various clients. With the package oriented design, the
insertion of custom interface and the interoperable use
of the resulting XML data structure the module is ca-
pable for other purposes or software projects.
4.2 Visualization
The Visualization of hierarchical structures like on-
tologies and the resulting large depth and/or width
often consumes to much space. Concerning to
Ben Shneiderman’s Informations Seeking Mantra:
”Overview first, zoom and filter, then details on de-
mand” (Shneiderman, 1996) the browser’s view is di-
vided into two major parts to consider the dynamic
structure of a collaborative created ontology. The
treemap approach (Shneiderman, 1991) allows to get
a fast and good overview of deep hierarchies. The
Figure 1: Treemap View
optimal space alignment and the quick navigation to
deeper layers provide a good starting point to get a
first overview. To distinguish the model’s hierarchy,
each tree-depth gets the same color, whereas the color
differs only in a slight amount to it’s parent color.
The second visualization, the graph view, is related
to a complex hierarchy. Especially ontologies with
their associations, parent and child classes and at-
tributes needs an easy respectively self-explanatory
visualization. To understand the enormous complex-
ity a deeper look on the construct of an ontology is
required. An ontology according to the guidelines of
Figure 2: Graph of model with parent, childs and associa-
tions
the OWL-Format is structured into classes and indi-
viduals. A class represents a more generic description
of an object, it links attributes and associations to it-
self and provides the inherited properties of its parent.
On the other side an individual or instance represents
a real world object, that is constructed from it’s cor-
responding model. To achieve an easy to understand
representation, we developed a metaphor which bears
in mind the dependencies of the model-individual re-
lation. The chosen metaphor maps the inheritance of
the models as a fluid, which flows from one model
to its submodel. The models layout looks like a kind
of reservoir whose filling height indicates the number
of individuals belonging to this model. If we proceed
Figure 3: Individual View of Type ”Lecture”
with this metaphor an individual could be seen as a
part of the fluid, consequently this appears as a drop,
which will show up in a zoomed view of the reservoir.
Further the drops can be filtered by attributes and as-
sociations to achieve clusters with common proper-
ties. However the models attributes are shown as
a box inside the reservoir, connections from this box
to other models indicate the related associations. To
keep the overview simple, only the current selected
model, its direct parent and child models and the as-
sociated models are rendered.
ICEIS 2005 - INFORMATION SYSTEMS ANALYSIS AND SPECIFICATION
296
Figure 4: Filtering of drops by different properties and as-
sociations
Figure 5: Individual in connection with properties and other
individuals
4.3 Awareness
The understanding of how an ontology evolved to
a current state is of particular importance. On the
one hand side an involved author should see which
changes are made to a model to recognize its current
progress and what intention another editor’s change
belong to. On the other hand an important case is
to provide a quick overview for future users, who
participate the collaborative process at a later time.
Therefore changes need to keep logged and facilities
for asynchronous conversations are essential. Hav-
ing much experience with Wiki-Systems we decided
to adopt the approach of logging every changed ver-
sion to provide the ability to return to a previous state.
This technique also enables an author to fork a previ-
ous version for new editing. Another media function
we will adopt is the asynchronous discussion related
to a model. With this we hope to achieve an effect of
continuous exchange between the users with the re-
sult of new respectively renewed views and in finding
a consensus.
4.4 Usability
To achieve a good usability the graphical user inter-
face is evaluated at specified milestones by students,
which indeed will be the future users. The current
framework of the in section Visualization mentioned
views provides the following components.
Tabbed Views The treemap and the graph view are
separated in different tabs. The selected model in
either one of this views will also change the current
model in the other tab. So a fast switching without
a new search of the selected model accelerate the
browsing of the ontology.
Bookmarks Adopted from World Wide Web -
Browsers, the feature of setting a bookmark for a
particular model provides a quick tie up to a previ-
ous session.
History A history navigation enables the user to keep
track of the visited models in a current session.
Address Field Due to the hierarchical nature of an
ontology an omnipresent virtual path (similar to a
filesystem path) will be printed to get informations
about the current location of a model. Further the
models are selectable which could also be used for
navigation purposes.
Zoomed Views Because of the lossless magnifica-
tion of Scalable Vector Graphics, zoomable views
are supported for the treemap and graph view and
also a general magnification of the whole applica-
tion which is especially useful for visually handi-
capped persons.
Directly Editable The editing is provided directly
at the specified model. Symbols like a drop or
an reservoir beneath the model indicate the result-
ing actions. The edit-forms itself are provided in
HTML, due to the text input limitations of SVG. To
create new associations a special copy stack which
holds the selected models will be initialized over
the whole session.
Related Filters As mentioned in section Visualiza-
tion, a model related filter will be provided to clus-
ter the individuals by specific attributes.
Automatic Linking Similar to the behavior of links
in Wiki-Systems, special types of links will han-
dled differently. Common WWW-links will be
considered and a direct jump to the page will be
performed. Links containing targets to pictures will
be rendered as picture instead of the textual appear-
ance. Special links to
open
sTeam items like users,
documents, groups and rooms can be entered sim-
ply by its unique id instead of a complicated URL.
Further it is planned to support the embedding of
audible and visual documents.
COLLABORATIVE ONTOLOGIES AND ITS VISUALISATION IN CSCW SYSTEMS
297
4.5 Participation
What we want to achieve is acceptance and participa-
tion for the cooperative creation of an ontology. The
challenge is to communicate the future user the ad-
vantage and gain for his own purposes. In case of on-
tologies stimuli could be to point out that the users in-
dividual knowledge will stay in conjunction to others
knowledge or that personal artifacts like Documents
can be embedded in a semantic structure to be useful
for others and in the end for the user himself. It is
surely not enough to just communicate it with words,
than with a direct response of the whole system. Fur-
ther attendance could be achieved by embedding the
system in university lectures or other workflow’s in
organizations. Especially organizations could profit
from sharing the knowledge of its members to keep it
beyond the individuals quitting. But covering this in
detail would go to far for this paper.
5 USAGE SCENARIOS
In this section an imaginable usage scenario could
take place. The
open
sTeam is available for all faculties
of the University of Paderborn. To mention a few, the
range of faculties goes from culture studies to com-
puter science over economics up to natural science.
Faculties outside the computer science uses the sys-
tem often for accompanying lectures of a semester.
After a semester the rooms and groups more or less
gets orphaned. The chances to create an ontology at
an ongoing lecture are obvious. Either for following
groups of the same faculty or for the chance of a real
interdisciplinary study, a group related or global on-
tology could help to keep the knowledge persistent
in the system, which in turn could help students to
revise on present semantic linked documents or re-
lations. For clarity, the following fictive example is
given.
5.1 Connecting knowledge domains
Imagine three courses, which will be held with sup-
port of the
open
sTeam environment. The first courses
topic is about how to write screenplays for documen-
tations, the second deals with current development of
robotics at the university and the third is about guide-
lines of university marketing. Now the three courses
create independently from each other models for their
topics in the global ontology. They embed documents
related to a specific field, relate these fields and doc-
uments to the involved students and to other previous
existing models of the ontology.
In the next semester an economic course should
create a short movie about the current research of
advanced technology at the university, which should
be presented to current and future investors. Now
the three above mentioned independent courses could
help to access related informations in a quick way.
Browsing the ontology to get informations about lead-
ings technologies and how to present them in attrac-
tive and for investors adequate form will help to keep
the focus on the real problems. Due to the seman-
tic structure of the ontology the students can reason
what documents are related for their tasks, who could
be a possible contact person and what else is associ-
ated. With the experience related to the creation of the
movie new associations could be developed, which
possibly connects the field of robotic with the field
of university marketing with the one of how to write
a screenplay.
This example should be more or less assignable to
any kind of organization which has either a broad or
a narrow spectrum of fields. Because of the ability
of every involved person to embed its experience and
knowledge, the approach in resolving a problem with
an already solved problem and the direct access of
related informations are the big chance of an ontol-
ogy. Facts like the constructive fashion and the never
reached final state of an ontology makes the essential
difference of a collaborative created ontology in con-
trast to an expert only one.
5.2 Mapping
open
sTeam structure
Another possible scenario is the automatic mapping
of the
open
sTeam structure to an ontology. The inter-
connection of users, room and documents could be vi-
sualized in a more detailed and clearly arranged man-
ner. The addition of a special type of model could
initiate a pre defined action, that performs changes in
both, the ontology and the system structure. A partic-
ular example in this case could be the administration
of the users room privileges or the relation of docu-
ments to specific rooms.
6 RELATED WORKS
One of the inspirations for visCOntE were on the one
hand Wiki Systems like its well known representa-
tive wikipedia
1
. Structures like the collaborative edit-
ing of linkable content as a main approach in vis-
COntE and other media elements like an asynchro-
nous item related discussion or the different handling
of URIs (like weblinks, images or email-addresses )
are adopted. Moreover the versioning of the Ontol-
ogy is much the same like it is in a Wiki - System,
visCOntE even extends this technique with the ability
1
http://www.wikipedia.org
ICEIS 2005 - INFORMATION SYSTEMS ANALYSIS AND SPECIFICATION
298
to work with a forked version of the ontology. On
the other hand ontology editors like Prot
´
eg
´
e
2
and
OntoEdit
c
3
with their fine granulated form based ap-
proach of ontology creation were taken into account
for the development. With the visual approach of on-
tology creation and the design goal to be a tool for
non experts visCOntE differs from these editors.
7 CONCLUSION AND FURTHER
WORK
Because of the independent format of the ontology,
the adaption is not limited to the context of the browse
and edit environment. Where visCOntE is the first
step in creating, searching and editing of the ontol-
ogy, other services could also operate on the provided
informations of it. This could be either in the way of
an automatic editing of the ontology by special rea-
sonings or in an automatic support of user interests.
General services can operate on any kind of ontol-
ogy regardless to the real information. An imagin-
able service or agent could observe a special model
of the ontology which a user places to a list of sub-
scribed interest. On any change he/she will be in-
formed about the current addition. Related to the us-
age scenario this could be a new document dealing
about robotics or a newly associated field. This could
help, to keep the user’s knowledge updated. A dif-
ferent service but with the same approach could be
the detection of people, who have similar interests
or problems. Especially in organization like univer-
sities this could help out to find new learning part-
ners. Another approach could be the initiation of a
keyword based semantical search engine that in fu-
ture could be embedded in visCOntE. The mentioned
scope is directed to the informations an ontology pro-
vides. Other services like a reminder of appointments
can operate only if the ontology has a relation to time.
To initiate these kind of services a special architec-
ture which analyses the attributes of an ontology is
needed. The OWL-Standard supports the assignment
of special data types like dates but others are speci-
fied by the author of the ontology. The special struc-
ture of the OWL-Format in separating datatypes from
the actual models makes it possible to provide such
architecture, that dynamically watches attributes and
relates them to special events. Perhaps, even a cre-
ation of services on the user side is possible to con-
struct a set of rules with corresponding actions. The
list could be widened by more particular examples.
What we wanted to mention is the wide field of appli-
cation of an ontology thats reasonably needs at first
2
http://protege.stanford.edu
3
http://www.ontoedit.de
to be constructed and in particular in a cooperative
way. Without this very first step, services like the
above remain meaningless, because the user is con-
strained to embed himself in the whole process. The
visContE project is the very first step to integrate a
new approach of the semantic web into the
open
sTeam
environment. Further steps could be, like mentioned
above, the integration of semantic webservices, the
sharing of the constructed ontology with other par-
ties or the representation of system characteristics or
other hierarchical based informations. Beside this, we
are looking forward to get an overview on social as-
pects and patterns of the systems user composition to
answer questions like which research fields are more
distinctive than others or how the cooperation of in-
terdisciplinary groups is possible / impossible. The
development of visContE is still in an early phase
of userinterface creation and implementation of the
server-module.
4
REFERENCES
Hampel, T. and Keil-Slawik, R. (2002). sTeam: Structuring
information in a team - distributed knowledge man-
agement in cooperative learning environments. ACM
Journal of Educational Resources in Computing Vol.
1, No. 2 (2002), 1-27.
Michael K. Smith, Chris Welty, D. L. M. (2004). Owl web
ontology language guide. http://www.w3.org/
TR/2004/REC-owl-guide-20040210/.
Ora Lassila, R. R. S. (1999). Resource descrip-
tion framework (rdf) model and syntax spec-
ification. http://www.w3.org/TR/1999/
REC-rdf-syntax-19990222.
Shneiderman, B. (1991). Tree visualization with tree-maps:
a 2-d space-filling approach. ACM Transaction on
Graphics , Vol. 11, No. 1 (1991), 92-99.
Shneiderman, B. (1996). The eyes have it: a task by data
type taxonomy for information visualisation. In Pro-
ceedings of the 1996 IEEE Symposium on Visual Lan-
guage (VL’96), 336-343.
Tim Berners-Lee, James Hendler, O. L. (2001).
The semantic web. Scientific American
http://www.scientificamerican.
com/article.cfm?articleID=
00048144-10D2-1C70-84A9809EC588EF21&catID=
2.
W3C (2004a). Semantic web. http://www.w3c.org/
2001/sw/.
W3C (2004b). Web ontology language (owl). http://
www.w3.org/2004/OWL/.
4
Visit http://steam.uni-paderborn.de/visconte/
COLLABORATIVE ONTOLOGIES AND ITS VISUALISATION IN CSCW SYSTEMS
299