A PRACTICAL ONTOLOGY-DRIVEN WORKFLOW
COMPOSITION FRAMEWORK
Huy Pham, Deborah Stacey and Rozita Dara
School of Computer Science, University of Guelph, Guelph, Canada
Keywords:
Workflow composition, Ontology-driven planning framework, Planning ontology, Ontology-driven workflow
composition, Workflow composition.
Abstract:
Existing planning-based approaches to ontology-driven workflow composition (ODWC) integrate planning
into their frameworks and ontologies in ways that are either less reusable or ineffective. A more modular and
reusable design is possible, but a successfull application of this design requires addressing some important
practicality issues. In this paper, we provide a comprehensive survey of existing approaches to intelligent
ontology-driven workflow composition, discuss the practicality challenges associated with building ontology-
driven and planning-based workflow composition systems, and present a reusable and practical ontology-
driven planning framework that could be used for such purposes.
1 INTRODUCTION
There are many application domains where a major
activity is the creation or arranging of sequences of
actions or events. This activity of temporal planning
is often referred to as workflow composition. Busi-
ness processes are often conceptualized as a workflow
and thus tools for helping with workflow composition
will find a wide and appreciative audience.
From a computational perspective, workflow com-
position is the task of coming up with an appropriate
sequence of (computational) tasks that accomplish a
given set of objectives. A simple example is the pro-
cess that a data analyst may go through when faced
with a data mining task. The analyst may employ
several data cleaning programs to make sure that the
data is consistent and in the proper format. They will
then use these data sets as input to a series of different
analysis programs and then send the output of these
programs to a visualization application to display the
results in a format accessible for business analysts.
Recent work in this area has taken advantage
of ontologies to drive this composition process.
But many existing approaches to planning-based
Ontology-Driven Workflow Composition (ODWC)
are not totally adequate for building effective real-
world workflow composition systems. Our analysis
leads us to believe that there are two major gaps that
need to be addressed in going from a domain expert
with a problem to solve and a system that allows that
expert to translate their domain problem into a com-
putational/computing problem that can generate a de-
sign for an appropriate workflow composition.
In this paper, we will provide a comprehensive
survey of existing approaches to intelligent ontology-
driven workflow composition and classify them into
four major approaches. We will then discuss the
challenges associated with building ontology-driven
and planning-based workflow composition systems
and identify features that are desirable in an effec-
tive ODWC framework. We then present our reusable
and practical ontology-driven planning framework
that incorporates all these features to provide the
ODWC system designer with a practical, effective
and friendly means of building ODWC systems. This
framework also addresses the important theoretical
challenges that are inherent to the task of doing goal-
based reasoning in ontology-driven applications. We
hope that our framework functions as a practical way
to address this challenge using existing and mature
technologies.
2 ONTOLOGY-DRIVEN
WORKFLOW COMPOSITION
In this section, we start out by reviewing some of the
most notable works in ODWC and make some obser-
vations about their advantages and limitations. We
242
Pham H., Stacey D. and Dara R..
A PRACTICAL ONTOLOGY-DRIVEN WORKFLOW COMPOSITION FRAMEWORK.
DOI: 10.5220/0003659102420248
In Proceedings of the International Conference on Knowledge Engineering and Ontology Development (KEOD-2011), pages 242-248
ISBN: 978-989-8425-80-5
Copyright
c
2011 SCITEPRESS (Science and Technology Publications, Lda.)
then place ourselves in the shoes of an ODWC sys-
tem designer and explore the list of features that are
desirable and important.
2.1 Existing Works
From an AI perspective, workflow composition is
a form of sequential decision making (Russell and
Norvig, 2002), and as such can be approached us-
ing various existing knowledge-based planning tech-
niques (Brachman and Levesque, 2004). The main
drawback of such an approach however, is that it re-
quires the knowledge bases – the collection of domain
knowledge needed to support the workflow composi-
tion process to be implicitly encoded in a language
that is specific to the planning framework employed,
and therefore less reusable to other applications (Gib-
son and Stevens, 2009).
In light of this reusability concern, the Semantic
Web community has been promoting the use of on-
tologies (Horrocks, 2008) as its key knowledge rep-
resentation mechanism. This mechanism provides
a standardized and well-supported facility that al-
lows domain knowledge to be described in a generic,
application-independent way that it can be easily un-
derstood, shared and reused by many different appli-
cations, including future applications for which the
knowledge was not originally intended.
Because of this important advantage that ontolo-
gies offer, there has been a lot of interest in repre-
senting the knowledge bases of workflow composi-
tion systems using ontologies instead of framework-
specific and application-specific languages, and con-
sequently, a large number of works have been re-
ported on ontology-driven workflow and system com-
position. The remaining part of this section provides
a review of some of the most notable works, as well
as some general observations about the common lim-
itations found in these works.
Most existing works in ODWC use the same over-
all approach: they wrap the various workflow ac-
tivities into logical units, called compositional units
(CU), adorn each CU with an appropriate set of at-
tributes (e.g. name, function, input data format, etc.),
organize them into some hierarchical structure and
serve them up via an ontology as a set of build-
ing blocks from which a workflow can be constructed.
Where they differ is the mechanism through which
each of these frameworks produce their workflows. In
this respect, existing works on ODWC can be roughly
divided into four different approaches.
2.1.1 Interactive Composition
In this approach, workflows are composed in an inter-
active fashion using manual inputs from the user. The
software composition framework described in (Hlo-
mani and Stacey, 2009) is one such example. This
framework aims to provide the foundations needed to
enable non-technical domain experts to rapidly com-
pose and modify their own software systems from
individual building blocks that represent the various
computing algorithms. More specifically, the frame-
work uses an ontology to formally describe the vari-
ous computing algorithms, and offers them as a set of
building blocks from which the user can visually com-
pose her computational system by selecting and con-
necting individual algorithmic components together.
One of the main limitations of this approach is that
while intelligent assistance is provided, the user is still
required to come up with the design.
2.1.2 Template-based Composition
In the second approach, workflows are selected from
a library of pre-built workflow templates and sug-
gested to the user as a guideline for composing her
system. This approach is represented by the data
mining workflow composition framework described
in (Morik and Scholz, 2003). The primary rationale
behind this approach is that successful and effective
workflows, once built, tested and tuned by experts,
could, and should, be shared and reused as guiding
templates for non-expert users in similar applications.
The main advantage of this approach is that such a
framework can act as a facility for easily capturing,
sharing and reusing effective data mining workflows.
The main limitation is that user assistance can only be
provided for situations in which a similar and appro-
priate template can be found in the library.
2.1.3 Planning-based Composition via Direct
Translation
In this approach, workflow composition is treated as
a planning problem and a planning mechanism is em-
ployed to generate workflows. Some of the most no-
table approaches are described in (Bernstein et al.,
2005), (
ˇ
Z
´
akov
´
a et al., 2008) and (Diamantini et al.,
2009). The main draw back of this approach is that
typically all planning-related knowledge is implic-
itly encoded into the primary knowledge base (i.e.
the ontologies used to describe the compositional
units). Because planning-related knowledge is appli-
cation and problem specific, embedding them into the
main ontologies makes these ontologies application-
A PRACTICAL ONTOLOGY-DRIVEN WORKFLOW COMPOSITION FRAMEWORK
243
and purpose-dependent and, as a consequence, less
reusable for future applications.
2.1.4 Planning-based Composition via a
Planning Ontology
This approach is represented by the works described
in (Gil et al., 2000) and (Rajpathak and Motta,
2004). These papers promote the use of a dedicated
and generic Planning Ontology to describe planning-
related knowledge.
In (Gil et al., 2000), the authors describe an ontol-
ogy that contains language constructs for describing,
among other things:
Objectives, subobjectives, and objectives decom-
position
Tasks (actions), task constraints, and tasks-
accomplish-objectives types of relationships
Decision points, ordering of tasks, temporal con-
straints
The strong point about this ontology is that the list
of language constructs it provides is comprehensive
and powerful enough to describes complex plans. The
main limitation of this work is that because its main
intent was to describe plans as opposed to planning
problems (the first is a solution of the second), the on-
tology lacks the necessary constructs (pre-conditions,
effects, world’s dynamics, cost/reward, etc.) needed
to drive a planner, and therefore is inadequate for the
purpose of building a planning-based ODWC system.
2.2 Desirable Features of an ODWC
Framework
In this section, we place ourselves in the shoes of
the ODWC system designer and explore the list of
features that are important or desirable for ODWC
systems. We will do this by trying to understand
the types of support the system designer would need
in order to build an effective and practical planning-
based ODWC system.
On the architectural side, the system designer
needs a framework that allows him to create simple
and reusable knowledge bases, as well as a framework
that can be employed to build WC systems in differ-
ent domains without extensive modifications or cus-
tom coding. On the representation side, he will need a
comprehensive set of language constructs for describ-
ing real-world workflow composition problems. On
the reasoning side, he will need a practical mecha-
nism to do goal-based reasoning in his system.
2.2.1 Clear Separation between Planning
Knowledge and other Domain Knowledge
This approach offers at least 3 important advantages
over previous approaches to planning-based ODWC.
1. Increased reusability of domain knowledge: Be-
cause the domain and objectives ontologies are
independent from the application that uses them,
they are much more reusable for other applica-
tions.
2. Easier to build descriptive domain ontologies:
Because the ontology designers can focus on de-
scribing the domains without having to worry
about how those descriptions will be used later,
high quality and descriptive domain ontologies
become easier to build.
3. Increased reusability and sharability of planning
knowledge and strategies: Because the syntax and
language constructs would allow planning prob-
lems to be declaratively described in a manner
that is completely independent from the imple-
mentation of the workflow composition engine,
the Planning Ontology has the potential to act as
a common abstraction language in which plan-
ning problems and strategies can be encoded in
a sharable and reusable way.
2.2.2 Planning Knowledge should be Captured
using an Ontology instead of Directly in
the Underlying Planning Formalism
The use of a planning ontology would allow the work-
flow composition problem, and the logic for how to
solve it, to be modelled independently from the un-
derlying planning paradigm. This separation is a very
important feature because it would allow the under-
lying translation framework to be reused for building
other ODWC systems without requiring a major mod-
ification to its logic. All that needs to be modified
instead is the ontological problem description.
2.2.3 Should provide a Comprehensive Set of
Language Constructs to Effectively
describe Real-world Workflow
Composition Problems
As discussed earlier, automated workflow composi-
tion, from a computational point of view, can be
considered a planning problem (Russell and Norvig,
2002) and, as such, can be modelled or described in
pretty much the same way as a planning problem.
A typical workflow composition problem can be ad-
equately modelled using the following types of de-
scriptions:
KEOD 2011 - International Conference on Knowledge Engineering and Ontology Development
244
Workflow Actions
Workflow Actions Pre-Conditions
Workflow Actions Effects (aka, Post-Conditions)
Initial State
Goal State
Once the above model has been built, it can be fed
into the workflow composition engine (WCE), which
searches through the many possible combinations of
action sequences and comes up with an appropriate
workflow that solves the problem.
In order to be able to build effective ODWC sys-
tems that can handle complex, real-world problems,
the system designer will need a comprehensive set of
language constructs that can effectively describe, at a
minimum, all the knowledge types above. While ex-
isting ontology-driven planning frameworks do pro-
vide some of these constructs, they still lack some
very important ones, including those needed to de-
scribe the world’s states (e.g. initial state, goal state),
state transitions, as well as those needed for advanced
planning techniques such as heuristics.
2.2.4 Should provide a Practical Mechanism for
doing Goal-based Reasoning in ODWC
Applications
From a knowledge representation and reasoning per-
spective, workflow composition boils down to the task
of proving the following entailment:
KB w Valid(w) AchieveGoal(w)
where w = [a1, a2, . . . , aN] is the workflow the user
is interested in, and KB is the knowledge base repre-
senting the workflow composition problem.
While several other knowledge-based formalisms
provide native support for this kind of reasoning, it
is not easily attainable in an ontology-driven appli-
cation. Based closely on Description Logics (DL),
ontologies were intended as a formalism for describ-
ing (object-oriented) worlds and is not well-suited for
goal-based reasoning tasks such as this.
First Order Logic (FOL) is the most expressive
language compared to DL and Horn Logic (HL) but
is undecidable. DL and HL are less expressive yet de-
cidable subsets of FOL. These two languages can be
thought of as two different ways of attaining decid-
ability via limiting the scope of the language. Due
to the difference in the ways their scopes are lim-
ited, each of these two languages are well-suited for
different purposes DL is well-suited for describing
(object-oriented) worlds, while HL is well-suited for
goal-based reasoning and hence problem solving.
While several approaches have been proposed to
integrate goal-based reasoning (e.g. Rules) into De-
scription Logics (c.f. the Semantic Web Rule Lan-
guage (SWRL) (Horrocks et al., 2004), and Descrip-
tion Logic Programs (DLP) (Grosof et al., 2003)),
their success to date is still somewhat limited. As
discussed in (Hitzler and Parsia, 2009), this is a non-
trivial task. The major challenge is that modifying or
extending the underlying language (DL in this case)
could lead to undecidability. SWRL, for example, is
undecidable and, as a consequence, has no native rea-
soner. Even more importantly, introducing a new ex-
tension to a language often entails some difficult and
time-consuming tasks. First, all the theoretical results
would need to be carefully mapped out and secondly,
the toolset will be need to be updated/augmented.
Thirdly, getting users to accept the new language is
also a highly non-trivial task. It usually takes a long
time for a language or formalism to attain the needed
critical mass for widespread adoption.
Due to these inherent difficulties, hybrid ap-
proaches are often considered when goal-based rea-
soning is needed within an ontology-driven applica-
tion. In section 3 below, we propose a hybrid (i.e.
translational) solution for integrating goal-based rea-
soning into ODWC systems.
3 A PRACTICAL ODWC
FRAMEWORK
3.1 The Overall Architecture
Figure 1: A generic architecture for planning-based
workflow composition systems. The planning ontology
and workflow composition engine constitute the proposed
framework and are reusable across domains and applica-
tions.
In our proposed architecture, as illustrated in Fig-
ure 1, a workflow composition system would utilize
three different ontologies to come up with its work-
flows.
A PRACTICAL ONTOLOGY-DRIVEN WORKFLOW COMPOSITION FRAMEWORK
245
First, an Objectives Ontology is used to describe
all the workflow objectives that the system is expected
to accept as its inputs. In a software project manage-
ment ODWC system, for instance, this objective on-
tology is used to describe objectives such as “Produce
a workflow that would minimize bug count” or “Pro-
duce a workflow that would minimize time to mar-
ket”, etc.
Similarly, an Operators Ontology will be used to
describe all the compositional units from which the
workflow can be composed. In a data mining ODWC
system, for example, the Operator KB will contain
the descriptions for all the different data mining and
preprocessing algorithms.
Finally, all planning-related knowledge needed to
drive the planner and build the workflow from the in-
dividual compositional units is captured in the plan-
ning KB. This KB is described by a domain and
reusable Planning Ontology. As discussed in sec-
tion 2.2.3 above, this ontology provides all the nec-
essary language constructs needed to describe work-
flow composition problems. A simplified version of
this ontology is shown in Figure 2.
Figure 2: A simplified view of the Planning Ontology show-
ing basic planning constructs such as states and actions.
It is worth noting that this architecture incorpo-
rates two of the desirable features identified in sec-
tion 2.2 above. Namely, it maintains a clear separa-
tion between planning-related knowledge and other
domain knowledge allowing the domain knowledge
to stay purpose-independent and hence more reusable
for future applications. It also allows the planning
logics to be described and modelled in an ontology
instead of directly in the underlying planning formal-
ism. As discussed earlier, this design feature allows
the framework to stay application-independent and
can be reused to build different workflow composi-
tion systems for different domains.
3.2 Integrating Goal-based Reasoning
into Ontology-driven Applications
As discussed in section 2.2.4 above, ontologies and
DLs do not provide native support for goal-based rea-
soning. We have taken a translational approach in
which the planning problem, once it has been de-
scribed by the ODWC system designer using the Plan-
ning Ontology, is translated into an equivalent ex-
ecutable program in HL and then executed using a
Prolog-based planning engine.
3.2.1 Model Closure
One of the standard cautions one has to take when in-
tegrating goal-based reasoning with ontological mod-
elling is the open-world vs closed-world assumptions
conflict. Goal-based program’s KBs are closed-world
models (a fact can be assumed to be false if it has
not been stated otherwise), while ontological mod-
els are open-world models (a fact cannot be assumed
false unless it was explicitly asserted). Currently how-
ever, this is not a practical issue in our situation – the
ODWC designer just has to picture a closed world in
his mind when describing his planning problem us-
ing the Planning Ontology. To help make it mentally
explicit for the designer however, we provide onto-
logical constructs that he can optionally insert into
the KB to logically (i.e. mentally) “close” the model
down. The statement “No other workflow actions are
available”, for example, when inserted into the KB,
has the effect of finalizing the list of already asserted
workflow actions and hence provides a mental closure
to the model.
3.2.2 Ensuring Translatability
Also, as with other translational approaches, one of
the main theoretical questions that is of importance
to our proposed approach is that of translatability
how does one ensure that the planning problem de-
scriptions created by the ODWC system designer are
always translatable into an executable planning pro-
gram in HL? To answer this question, a few observa-
tions are in order.
First, a well-defined ontology can be thought of as
a form of language the list of concepts it provides
constitutes the vocabulary of the language, while the
roles it defines dictates the ways in which the vocab-
ulary can be combined together to form statements.
Secondly, by carefully controlling the list of concepts
and roles in the ontology, we can restrict or control
the types of statements one can express using the on-
tology.
With these observations in mind, one can see that
by being very selective and careful with the language
constructs in the Planning Ontology, we can ensure
that all possible workflow composition problem de-
scriptions are translatable to executable planning pro-
grams in HL. This, in fact, is the main intuition behind
our approach. Figure 3 provide a visual illustration
KEOD 2011 - International Conference on Knowledge Engineering and Ontology Development
246
for this intuition.
Figure 3: The primary intuition behind our proposed ap-
proach: the Planning Ontology acts as a restrainer that helps
ensure the description of the planning problem always falls
into a sub area of Description Logics that is translatable to
an executable program in Horn Logic
3.2.3 Practical Benefits
By taking a translational approach, as opposed to
a language modification or extension approach, our
framework is able to make use of an existing and
mature theoretical framework (HL) and technologies
(Prolog) to provide seamless goal-based reasoning ca-
pability in ontology-driven applications.
Another advantage offered by our framework is
that it allows the ODWC system designer to continue
to think and work in the ontological modelling envi-
ronment that he or she is already comfortable with,
without having to learn a new language extension or
a new planning formalism.
3.3 Addressing Effectiveness Issues
3.3.1 Incorporating the System Designer’s
Heuristic Advice
Real-world workflow composition problems often re-
sult in highly complex planning problems that are
well beyond the capabilities of even the most ad-
vanced automated planners. Fortunately, in almost
all cases, the ODWC framework designer will have
some valuable procedural insights on how to take ad-
vantage of the problem’s structure or how to best go
about solving the problem. An effective and practical
planning-based ODWC framework, therefore, must
provide convenient facilities and mechanisms for the
designer to provide these insights to his system. In
our framework, this can be done via a mechanism
called partial programming (Reiter, 2001). The idea
is that, instead of relying solely on the planner to gen-
erate workflows from scratch, the workflow compo-
sition system would start from a partial template that
the system designer has provided. Because this tem-
plate contains all the heuristic advice from the de-
signer, computing a workflow from this template will
be much faster and efficient than computing one from
scratch.
Using our framework, the system designer creates
these partial programs (i.e. templates) using the lan-
guage constructs provided by the Planning Ontology.
We describes some of these language constructs be-
low:
Action Template. This concept is used to repre-
sent a subworkflow, and is the equivalence of a
procedure in a programming language. An action
template can contain workflow actions (i.e. com-
positional units) or other action templates. The
constituents of the template can be arranged in a
simple sequential order or in complex procedural
orders such as loops (while, foreach) or branch
(if..then..else).
ChoicePoint. This concept is used to represent a
nondeterministic choice between two or more Ac-
tionTemplates (i.e. subworkflows). The designer
would use this construct to convey to his work-
flow composition system that, instead of having to
consider all possibilities, it can narrow its choices
to just the subset of subworkflows specified in
the choice point. The more insight the designer
has about the workflow composition problem, the
more choice points he will put in the template and
the less work the system has to do.
ArgChoicePoint. Each workflow action might
take one or more input arguments and an Arg-
ChoicePoint can be used to narrow down the pos-
sible argument values the system has to consider.
3.3.2 Hierarchical Composition
Our Planning Ontology also provides language con-
structs for dealing with hierarchical workflow compo-
sition. In particular, the ODWC system designer can
use the MacroAction concept to annotate (or wrap) a
subworkflow into a compositional unit, with its own
pre-conditions and effects. The workflow composi-
tion system can then use these macro units, along-
side with other regular units in a seamless manner, to
quickly compute high-level workflows satisfying the
objective. Once such a workflow has been success-
fully found, the macro units can be iteratively fleshed
out into concrete subworkflows.
A PRACTICAL ONTOLOGY-DRIVEN WORKFLOW COMPOSITION FRAMEWORK
247
4 SUMMARY
In this paper, we reviewed the overall landscape
of ontology-driven workflow composition, and de-
scribed the four major approaches to which existing
works can be classified. We have also identified and
explained a list of features that are desirable for an
effective ODWC framework and proposed a practi-
cal framework that incorporates all these features to
provide the ODWC system designer with a practical,
effective and friendly means of building ODWC sys-
tems.
We have also provided a discussion on how impor-
tant theoretical challenges that are inherent to the task
of doing goal-based reasoning in ontology-driven ap-
plications can be practically addressed using existing
and mature technologies.
As for future works, we are working to provide
a formal proof on the translatability of workflow
composition problem descriptions. Additionally, are
also working to incorporate the notions of planning
cost/reward and concurrency into our planning frame-
work to allow it to produce non-linear workflows that
not only accomplish the give objective, but also ac-
complish it in an optimal way. Also, each workflow
action can have non-deterministic effects, and we are
also looking to incorporating non-determinism into
our framework.
A longer version of this paper including a case
study can be found at ontology.socs.uoguelph.ca.
REFERENCES
Bernstein, A., Provost, F., and Hill, S. (2005). Toward
intelligent assistance for a data mining process: An
ontology-based approach for cost-sensitive classifica-
tion. IEEE Transactions on Knowledge and Data En-
gineering, 17:503–518.
Brachman, R. and Levesque, H. (2004). Knowledge Repre-
sentation and Reasoning. Morgan Kaufmann Publish-
ers Inc., San Francisco, CA, USA.
Diamantini, C., Potena, D., and Storti, E. (2009). Ontology-
driven kdd process composition. In Proceedings
of the 8th International Symposium on Intelligent
Data Analysis: Advances in Intelligent Data Analy-
sis VIII, IDA ’09, pages 285–296, Berlin, Heidelberg.
Springer-Verlag.
Gibson, A. and Stevens, R. (2009). Introduction to ontolo-
gies. In Popescu, M. and Xu, D., editors, Data Min-
ing Applications Using Ontologies in Biomedicine.
Artech Hous.
Gil, Y., Gil, A., and Blythe, J. (2000). PLANET: A Share-
able and Reusable Ontology for Representing Plans.
In Proceedings of the AAAI Workshop on Representa-
tional Issues.
Grosof, B. N., Horrocks, I., Volz, R., and Decker, S. (2003).
Description logic programs: combining logic pro-
grams with description logic. In WWW ’03: Proceed-
ings of the 12th International Conference on World
Wide Web, pages 48–57, New York, NY, USA. ACM.
Hitzler, P. and Parsia, B. (2009). Ontologies and rules.
In Bernus, P., Blazewics, J., Schmidt, G., Shaw,
M., Staab, S., and Studer, R., editors, Handbook on
Ontologies, International Handbooks on Information
Systems, pages 111–132. Springer Berlin Heidelberg.
Hlomani, H. and Stacey, D. (2009). An ontology-driven ap-
proach to software system composition. In Proceed-
ings of KEOD 2009.
Horrocks, I. (2008). Ontologies and the Semantic Web.
Commun. ACM, 51(12):58–67.
Horrocks, I., Patel-Schneider, P. F., Boley, H., Tabet, S.,
Grosof, B., and Dean, M. (2004). SWRL: A Semantic
Web Rule Language Combining OWL and RuleML.
Technical report, World Wide Web Consortium.
Morik, K. and Scholz, M. (2003). The miningmart approach
to knowledge discovery in databases. In In Ning
Zhong and Jiming Liu, editors, Intelligent Technolo-
gies for Information Analysis, pages 47–65. Springer.
Rajpathak, D. and Motta, E. (2004). An Ontological For-
malization of the Planning Task. In Proceedings In-
ternational Conference on Formal Ontologies in In-
formation Systems (FOIS’04), Torino, Italy.
Reiter, R. (2001). Knowledge in Action: Logical Founda-
tions for Specifying and Implementing Dynamical Sys-
tems. The MIT Press, Massachusetts, MA, illustrated
edition edition.
Russell, S. J. and Norvig, P. (2002). Artificial Intelligence:
A Modern Approach (2nd Edition). Prentice Hall.
ˇ
Z
´
akov
´
a, M., K
ˇ
remen, P.,
ˇ
Zelezn
´
y, F., and Lavra
ˇ
c, N.
(2008). Using ontological reasoning and planning
for data mining workflow composition. In SoKD:
ECML/PKDD 2008 Workshop on Third Generation
Data Mining: Towards Service-Oriented Knowledge
Discovery.
KEOD 2011 - International Conference on Knowledge Engineering and Ontology Development
248