![](bg3.png)
From a more technical point of view, however,
these two types of reasoning are very different in na-
ture and, as a result, most knowledge representation
and reasoning formalisms typically support just one
type or the other. Traditionally, planning is supported
by rule-based formalisms that are based on Logic Pro-
gramming (e.g., Prolog) while ontological reasoning
are supported by object-oriented formalisms that are
based on Description Logics (e.g., Ontologies).
Many practical ontology-driven applications,
however, often require both types of reasoning and,
as a consequence, there has been a great deal of inter-
est in combining planning and ontological reasoning
into a single formalism. Unfortunately, as explained
in (Hitzler and Parsia, 2009), this is an inherently
non-trivial task. First, combining the two formalisms
leads to semantic-related issues because rule-based
formalisms typically assume a closed-world model
while description logics based formalisms assume an
open world model. Second, adding language features
from one formalism to the other often result in an un-
decidable language. In the next section, we briefly
describes how existing approaches cope with these
challenges, and discuss the pros and cons of each ap-
proach.
2.3 Existing Works on Integrating
Planning into Ontology-driven
Applications
Generally speaking, existing works on integrating
planning into ontology-driven applications can be di-
vided into three main approaches: Language Modi-
fication, Parallel Modelling, and Translation. In the
first approach, the underlying language (i.e., Descrip-
tion Logics) is modified or extended to support rule-
based reasoning. In the second approach, applica-
tion knowledge are described in ontologies, while
planning-related information are described separately
in a rule-based language. In the third approach,
planning-related knowledge are described using an
ontology, alongside with other application knowl-
edge, and translated into an executable rule-based
planning program. We describe these approaches in
more details below.
2.3.1 Language Modification Approaches
Because ontological reasoning is a feature of on-
tologies, and planning is a feature of rule-based for-
malisms, it is a fair question to ask if rules and on-
tologies can be reasonably combined to produce a
more or less unified language in which both rule-
based and ontological reasonings are dually supported
in a seamless way. A lot of work in this direction have
been reported, and readers who are interested in this
topic are refered to (Hitzler and Parsia, 2009) for an
overview, and (Horrocks et al., 2004), (Grosof et al.,
2003), and (Motik and Rosati, 2008) for some better
known example approaches. Here, we will focus our
discussion instead on the pros and cons of such an ap-
proach.
Briefly speaking, the main advantage of a lan-
guage modification approach is that of theoretical ele-
gance. If successful, such a framework can serve as a
unifying formalism that combines features from both
rules and ontologies, two well-established knowledge
representation and reasoning formalisms. The main
disadvantage of this approach, however, is that it is
inherently difficult, and success has been very lim-
ited so far (Hitzler and Parsia, 2009). In addition to
the semantic (i.e., open world vs closed world) and
complexity (i.e., decidable reasoning algorithms) is-
sues mentioned in the previous section, modifying or
extending a language often entails several other im-
portant tasks. First, adequate tooling support will
need to be provided for the new language. This in-
cludes efficient reasoner implementations (assuming
the new language is decidable) and effective editors
for authoring models in the new language. Second,
adequate experience reports will also have to be pro-
vided. This includes, among other things, case studies
showing how such a formalism can be applied to solve
practical real-world problems.
Given the difficult theoretical challenges above,
and given the fact that most works in this direction are
still in their early stages, it can be seen that language-
based approaches, while theoretically rewarding and
important, also have some disadvantages when con-
sidered as a mean for bringing planning to ontology-
driven applications.
2.3.2 Parallel Modelling Approaches
Another popular approach to bringing planning to
ontology-driven applications is the “parallel” ap-
proach in which planning and application knowledge
are kept separated in two parallel worlds: planning-
related knowledge are described in a (rule-based)
planning language, while other application knowl-
edge are described in ontologies. Integration is done
by querying the ontologies for the list of available
planning actions, and perhaps their pre-conditions,
and executing the planning program using these ac-
tions.
Several works from the ontology-driven workflow
composition community have been reported to follow
this approach. (Bernstein et al., 2005), (
ˇ
Z
´
akov
´
a et al.,
2008) and (Diamantini et al., 2009), for example, de-
PRACTICAL GOAL-BASED REASONING IN ONTOLOGY-DRIVEN APPLICATIONS
101