SEMANTIC ARGUMENTATION
IN DYNAMIC ENVIRONMENTS
J¨orn Sprado and Bj¨orn Gottfried
Centre for Computing Technologies, University of Bremen, Germany
Keywords:
Decision support, Argumentation systems, Ontologies, Spatiotemporal knowledge management.
Abstract:
Decision Support Systems play a crucial role when controversial points of views are to be considered in
order to make decisions. In this paper we outline a framework for argumentation and decision support. This
framework defines arguments which refer to conceptual descriptions of the given state of affairs. Based on
their meaning and based on preferences that adopt specific viewpoints, it is possible to determine consistent
positions depending on these viewpoints. We investigate our approach by examining soccer games, since
many observed spatiotemporal behaviours in soccer can be interpreted differently. Hence, the soccer domain
is particularly suitable for investigating spatiotemporal decision support systems.
1 INTRODUCTION
Over the last years there has been an increasing in-
terest in the fields of Decision Support Systems and
Artificial Intelligence (AI) to investigate argumenta-
tion approaches. As a promising model for reasoning
about inconsistent knowledge (Amgoud et al., 2008;
Bench-Capon and Dunne, 2007) we investigate argu-
mentation frameworks to examine the behaviours of
objects, i.e. we are looking for how to provide deci-
sion support in the context of spatiotemporal systems.
That is an example in the soccer domain looks like
this: from one point of view the behaviours of soccer
player might be inconsistent since not all behaviours
support a given strategy; however, another point of
view might argue for another strategy for which the
behaviours are consistent; and yet another view would
state that the players were not able to get to a common
strategy – it is then the decision of the coach which
player to censure in which way and with which kinds
of arguments. Argumentation frameworks allow spe-
cific conclusions to be derived about what is true or
rather forms a consistent argumentation. In our case
in order to evaluate spatiotemporal group interactions.
1.1 Motivation
Fig. 1 shows the 73rd minute of the game Costa Rica–
Germany of the world championship 2006: players
of the black team (Costa Rica) attack the white team
10
1
3
17
21
16
9
6
5
11
20
10
1
3
17
21
16
9
6
5
11
20
10
1
3
17
21
16
9
6
5
11
20
10
1
3
17
21
16
9
6
5
11
20
10
Figure 1: Four following scenes of the 73rd minute of
the game Costa Rica–Germany of the world championship
2006
(Germany); player no. 9 tries to run towards the front
middle in order to receive the ball the attacker is going
to pass through the German players. The motion pat-
tern among the teammates of Costa Rica is a pattern in
which all players move towards the goal. Wanchope
(no. 9) runs into the penalty area, preparing for getting
the pass from no. 10. The pattern among the German
players shows the tendency of two of the teammates
to meet in order to avoid a gap and to get the ball. That
is Metzelder (no. 21) and Mertesacker (no. 17) move
towards the attacker while Friedrich (no. 3) moves to-
wards the middle expecting the ball in the middle, try-
ing to beat Wanchope. Comparing these two patterns
within the teams it clearly shows that the strategy of
236
Sprado J. and Gottfried B. (2009).
SEMANTIC ARGUMENTATION IN DYNAMIC ENVIRONMENTS.
In Proceedings of the 11th International Conference on Enterprise Information Systems - Artificial Intelligence and Decision Support Systems, pages
236-241
DOI: 10.5220/0002010702360241
Copyright
c
 SciTePress

the Costa Rica teammates follow one strategy, while
the German defenders follow simultaneously two dif-
ferent strategies: no. 21 and no. 17 fight the attacker,
while no. 3 prepares for dealing with no. 20 of the
other team. The decision of no. 3 to keep in touch
with his opposing player spoils a common strategy of
the German team which could be an offside trap.
The described scene illustrates a complex situa-
tion where different objects interact in a dynamic en-
vironment. To argue in such a complex situation we
have to consider all those interactions.
1.2 Overview
A typical argumentation generation process is shown
in Fig. 2. Input of the whole process are raw po-
Sec. 5
Sec. 4
Raw Positional
Data
1
Arguments
2
Semantic
Arguments
5
Preferred
Extensions
8
Explanation
Component
9
Attack-Relations
3
Validity Intervals
6
Terminology
4
Audiences
7
Figure 2: Argumentation generation as a process.
sitional data and the output is an explanation com-
ponent for the input data. The input data represents
at the most basic level spatiotemporal object interac-
tions. Conversely, an explanation of such object in-
teractions is provided by the explanation component.
An explanation describes at an abstract semantic level
what is going on, while the raw positional input data
are only positional measurements. A generated expla-
nation might in particular be useful for making deci-
sions. The complexity of such a process is due to the
goal of bridging the gap between pure measurements
and meaningful explanations of those measurements.
For this purpose basically two methods are employed:
• argumentation frameworks are used in order to
describe which concepts form consistent scenar-
ios, and conversely, which concepts cannot be rec-
onciled (Dung, 1995);
• the terminology of both a specific domain and of
argumentation frameworks is defined by methods
of description logics (Baader et al., 2003).
The process of explanation generation, and hence de-
cision support, is thus, a process of looking for con-
sistent sets of arguments. Domain specific arguments
are defined by an ontology, and the more abstract
level of argumentation frameworks is itself described
by another ontology. Before interpreting data, a pri-
ori knowledge is modelled by means of these on-
tologies. The process is as follows; taking the raw
positional data (Box 1), arguments about basic spa-
tiotemporal behaviour patterns are constructed (Box
2). Then attack-relations among arguments are con-
structed (Box 3) that define which arguments attack
other arguments. Referring to terminological knowl-
edge (Box 4) which is defined a priory, semantic ar-
guments (Box 5) are constructed; they describe at
the semantic level how concepts characterise the data.
Which arguments temporally relate are described by
validity intervals (Box 6): arguments might relate be-
cause they follow each other and can therefore in-
fluence each other; or they can temporally coincide,
or they can at least temporally overlap. Audiences
(Box 7) determine specific viewpoints which can be
taken in order to justify a specific argumentation.
Audiences influence which of the currently instanti-
ated arguments do form consistent sets of arguments
with respect to these audiences, that is preferences are
made (Box 8). Preferences and audiences, determine
eventually which spatiotemporal behaviours stand for
which arguments, and hence decisions, when taking a
specific point of view.
While in (Sprado and Gottfried, 2008b) the anal-
ysis of raw positional data is investigated and while
in (Sprado and Gottfried, 2008a) the argument con-
struction process is investigated, this paper focuses on
the argumentation process itself at the semantic level.
The explanation component is not further dealt with
here, but the computation results of preferred exten-
sions provide the core knowledge which is necessary
in order to develop explanation components.
After having revisited description logics (Sec. 2)
and argumentation frameworks (Sec. 3), in Sec. 4 we
propose our approach how to describe arguments se-
mantically by a given set of arguments and attack-
relations. Then in Sec. 5.1 audiences are introduced
on which consistent positions can be found (Sec. 5.2).
Finally, Sec. 6 concludes with a discussion.
2 ONTOLOGIES AND
DESCRIPTION LOGICS
A popular definition of ontologies in AI is proposed
by Gruber:
SEMANTIC ARGUMENTATION IN DYNAMIC ENVIRONMENTS
237

an ontology is an explicit formal specification
of a shared conceptualisation (Gruber, 1995).
Such a conceptualisation corresponds to a way of
thinking about some domain (Uschold, 1998). Fol-
lowing the W3C Web Ontology language (Patel-
Schneider et al., 2004) the ontologies shown in this
paper are expressed by using a description logic (DL).
Description logics are part of the family of knowledge
representation languages that are subsets of first-order
logic (cf. (Sattler et al., 2003)). A description logic is
a concept-based knowledge representation formalism
with well-defined model-theoretic semantics (Baader
et al., 2003). A DL consists of atomic concepts (unary
predicates), atomic roles (binary predicates), and in-
dividuals (constants). The expressive power of DL
languages is restricted to a small set of constructors
for building complex concepts and roles. Implicit
knowledge about concepts and individuals can be
inferred automatically through using inference pro-
cedures. For precise formal semantics of DLs see
(Baader et al., 2003).
A description logic knowledge base is naturally
separated into two parts: a TBox containing inten-
sional knowledge that describes general properties of
concepts and an ABox containing extensional knowl-
edge that is specific to the individuals of the universe
of discourse.
3 ARGUMENTATION
FRAMEWORKS
In this paper we use argumentation frameworks in
order to construct and compare arguments. An ab-
stract argumentation framework is proposed by Dung
(Dung, 1995). We stick accordingly to his formalisa-
tion:
Definition 1 (Argumentation Framework). An
argumentation framework (AF) is a pair AF =
(A R ,attack), where A R is a set of arguments and
attack ⊆ A R × A R is an attack-relation.
The relation attack(x,y) expresses that an argu-
ment x attacks another argument y. The acceptability
of arguments can be defined based on the notions of
defence and conflict-freeness.
Definition 2 (Conflict-freeness). Let S ⊆ A R be a
set of arguments in an argumentation framework AF.
A set of arguments S is conflict-free iff there are no
arguments x,y ∈ S with ∃x∃y attack(x,y).
Definition 3 (Defence). A set of arguments S de-
fends an argument x ∈ A R , that is to say x is accept-
able with respect to S denoted as acceptable(x,y), iff
∀y ∈ A R ∃z ∈ S attack(y,x) → attack(z,y).
Acceptability semantics havealso been introduced
in Dung’s abstract argumentationframework. In order
to find consistent positions within an AF (cf. Sec. 1)
preferred extensions are of interest.
Definition 4 (Preferred Extension). A conflict-free
set of arguments S is admissible if ∀x acceptable(x,S)
with x ∈ S . A conflict-free and admissible set of ar-
guments S is a preferred extension iff it is a maxi-
mal (with respect to set inclusion) admissible subset
of A R .
Furthermore, a value-based argumentation system
is defined by Bench-Capon (Bench-Capon, 2003) as
follows:
Definition 5 (Value-based AF). A value-based ar-
gumentation framework (VAF) is a triple VAF =
(H (A R ,attack)),V ,η), where (A R ,attack) is an
argumentation framework, V is a non-empty set of
values and η : A R → V is a function that associates
each argument x ∈ A R with a value η(x) ∈ V .
The set of values V within the VAF represents
types of arguments. Based on these values different
audiences enable the consideration of diverse posi-
tions.
Definition 6 (Audience). Let A R be a set of argu-
ments. An audience for a V A F is a binary relation
R ⊂ A R × A R whose (irreflexive) transitive closure
R
∗
is asymmetric.
Definition 7 (Acceptability Semantics).
Let (H (A R , attack)),V ,η) be a VAF and R an au-
dience.
1. A set of arguments S is conflict-freewith respect to
an audience R iff there are no arguments x,y ∈ S
with ∃x∃y attacks
R
(x,y).
2. An argument x is acceptable with respect to a
set of arguments S and an audience R denoted
as acceptable
R
(x,S ) iff ∀y∃z attacks
R
(y,x) →
attacks
R
(z,y) with x, y ∈ A R and z ∈ S .
3. A conflict-free set of arguments S is admissi-
ble with respect to an audience R denoted as
admissible
R
(S ) iff ∀x acceptable
R
(x,S ) with x ∈
S .
4. A set of arguments S is a preferred extension for
an audience R iff it is a maximal (with respect to
set inclusion) admissible subset of A R with re-
spect to an audience R .
4 SEMANTIC DESCRIPTION OF
ARGUMENTS
Arguments and their attack-relations have been ob-
tained by analysing raw positional data before de-
ICEIS 2009 - International Conference on Enterprise Information Systems
238

Argument (A)
Object
ArgumentType
BallPlayer
Defender Striker
istypeOf
Measurement
Quantitative
Measurement
Qualitative
Measurement
hasSource
AnalysisLevel
Strategy
Tactic
Action
Sweeper
hasAnalysisLevel
ArgumentNode
Conclusion
hasPremise
hasConclusion
1...*
1
Premise
ARGUMENTSOCCER
ZonalDefence
ManMarking
hasTargetActor
Location
hasLocation
Left-wing Right-wing
Penalty-area
Wing
...
...
...
...
Concept PropertyGeneralisation1...* Cardinality
hasInitialActor
Tactic
Argument
Action
Argument
Situation
Argument
ZonalDefence
Argument (ZDA)
DefensivePlay
Argument (DPA)
Strategy
Argument (ST)
OffensivePlay
Argument
...
...
...
...
...
ManMarking
Argument (MMA)
...
BallorientedZonalDefence
Argument (BZDA)
...
Figure 3: Concepts and relations of the soccer and argu-
ments domain ontology. For simplicity, only some of the
concepts and relations are shown.
scribing the meaning of arguments (cf. Fig. 2). Ta-
ble 1 presents some informal argumentsfor the scenes
shown in Fig. 1. For instance, argument A means that
player no. 20 of the black team is marked by no. 3 of
the white team. Argument B says that two players of
the white team attack no. 10 of the black team. More-
over, arguments C and D state different moving di-
rections of pairs of objects. Argument E points out an
offside-trap of the white team which is in conflict with
the arguments C, D and F. Eventually, arguments F
and G denote different strategies of the white team;
a man-to-man marking strategy is completely differ-
ent to a zonal defence strategy. Further conflicts can
be found among these arguments. Additionally, there
are different kinds of arguments, e.g. two objects are
involved in argument A which says something about
an action between players, while argument F makes a
statement about a team strategy.
The idea of a semantic description of arguments
is to represent their meaning explicitly and thus, to
exploit machine-processable metadata to arguments.
For this purpose we use ontologies which already
have been widely discussed in the Semantic Web area.
Arguments with clear semantics become machine-
interpretable and further inference mechanisms can
be applied (cf. large-scale argumentation (Rahwan,
2008; Rahwan and Banihashemi, 2008)).
In order to establish the link between the infor-
mal description of an argument and its ontological de-
scription we follow the idea of registration mappings
(Bowers et al., 2004) to have separate descriptions of
structural details and semantics. This has the advan-
tage that the semantics of an argument can be spec-
ified more accurately because the specification does
not try to mirror the structure of the arguments.
Table 1: Example arguments and attack-relations.
Id Meaning A
A man-marking of black no. 20 B, E, G
by white no. 3
B double-teaming of black no. 10 A, F
by white no. 17 and no. 21
C different moving direction E, G
of white no. 3 and no. 17
D different moving direction E, G
of white no. 3 and no. 21
E offside-trap of team white A, C, D
F man-to-man marking B,G
of team white
G zonal defence of team white A, C, D, F
5 CONCEPT-BASED
ARGUMENTATION
Our running example involves multiple arguments for
and against different claims. In this section we show
exemplarily how to reconcile these conflicts and how
to form consistent sets of arguments using our new
approach on concept-based argumentation.
Preferred extensions are of main interest because
they enable to form consistent sets of arguments.
While we look at preference-based argumentations in
dynamic scenes we focus on obtaining preferred ex-
tensions which hold with respect to an audience for an
observed time interval. We know that every argumen-
tation framework has at least one preferred extension
which might be the empty set (Dung, 1995). There-
fore, the empty set would be a solution in general.
But we are interested in finding out whether there are
other admissible sets of arguments.
We define concept-based argumentation frame-
works in the following way:
Definition 8 (Abstract Concept-based AF). A DL
knowledge base K = (T , A ) represents the domain
of interest, where T is a TBox and A is a ABox. A
concept-based argumentation framework is a triple
CAF = (H (A R , attack),A ,ν),
where H is an argumentation framework and
ν : A R → A is a mapping that assigns ABox asser-
tions to each argument y ∈ A R , denoted as x = ν(y).
The concept-based argumentation framework is a
specialisation of a VAF which associates concepts in-
stead of simple values with arguments. That is to
determine the acceptability of arguments we can re-
fer to VAFs (Bench-Capon, 2003). However, a main
SEMANTIC ARGUMENTATION IN DYNAMIC ENVIRONMENTS
239

difference is that semantic arguments are mapped to
concepts automatically as a result of inference pro-
cedures. For instance, if we change the semantic
description of an argument, an appropriate mapping
within the argumentation system will be automati-
cally adjusted. Thus, argumentation becomes more
flexible and more scalable.
5.1 Concept-based Audiences
According to VAFs (Bench-Capon, 2003) a concept-
based argumentation framework also provides mech-
anisms for considering diverse positions in one argu-
mentation system. We use audiences in order to deter-
mine multiple preferred extensions for different view-
points.
Definition 9 (Concept-based Audience). Let K =
(T ,A ) be a DL knowledge base, C be a set of con-
cepts of T and CAF = (H (A R , attack),A ,ν) be
an abstract concept-based argumentation framework.
An audience I for a C A F is a binary relation I ⊂
C × C whose taxonomic expansion is asymmetric. An
argument a
m
is preferred to a
n
in the audience I de-
noted (a
m
≻ a
n
) if (a
m
,a
n
) ∈ I
T
.
In contrast to Bench-Capon, we allow preferences
among arguments to be described with concepts in-
stead of simple values. We can define preferences on
any sub-concepts of an Argument. If we like to prefer
arguments (cf. Fig. 3) of type ZDA (ZonalDefenceAr-
gument) to those of MMA (ManMarkingArgument) this
can be denoted as ZDA ≻ MMA.
As we state preferences on concepts we have to
consider that preferences are also propagated to all
sub-concepts which will be ensured by a taxonomic
expansion. This means that knowledge in terms of
concept A subsumes concept B will be directly rep-
resented within an audience. Consequently, a taxo-
nomic expansion extends an audience with concepts
of the subsumption hierarchy.
Definition 10 (Taxonomic Expansion). Let K =
(T ,A ) be a DL knowledge base, C be a set of con-
cepts of T and I an appropriate audience. An audi-
ence is a taxonomic expansion I
T
of an audience I
with respect to a TBox T that satisfies the following
constraints for all possible pairs (C
n
,C
m
) of C :
If (C
n
,C
m
) ∈ I then (C
n
,C
m
) ∈ I
T
and
– if there is a sub-conceptC
sub1
⊑ C
n
with respect to
T for a pair (C
n
,C
m
) ∈ I then (C
sub1
,C
m
) ∈ I
T
,
– if there is a sub-concept C
sub2
⊑ C
m
with respect
to T for a pair (C
n
,C
m
) ∈ I then (C
n
,C
sub2
) ∈ I
T
,
– and if there are sub-concepts C
sub1
,C
sub2
with
C
sub1
⊑ C
n
and C
sub2
⊑ C
m
with respect to T for a
pair (C
n
,C
m
) ∈ I then (C
sub1
,C
sub2
) ∈ I .
In case of (C
n
,C
m
) /∈ I and C
n
⊑ C
m
wrt T :
1. if we mainly prefer more general concepts then
(C
m
,C
n
) ∈ I
T
. Such a taxonomic expansion will
be denoted as I
T >
.
2. otherwise if we mainly prefer more specific ones
then (C
n
,C
m
) ∈ I
T
. Such a taxonomic expansion
will be denoted as I
T <
.
Bench-Capon already mentioned for value-based
argumentation frameworks that an audience I typi-
cally does not describe a unique total ordering of the
values; there are in fact multiple compatible order-
ings. They are referred to as specific audiences com-
patible with I (Bench-Capon et al., 2007). This also
holds for concept-based audiences.
Definition 11 (Specific Audience). Let K = (T ,A )
be a DL knowledge base, C be a set of concepts and I
an audience. A specific audience α is a total ordering
of C with respect to an audience I and
∀C
1
,C
2
∈ C : hC
1
,C
2
i ∈ α =⇒ hC
2
,C
1
i /∈ I
T
In relation to Bench-Capon we denote a set of spe-
cific audiences with χ(I ).
Example 1. Let A, B, C, D and E be concepts and
B ⊑ A,C ⊑ A,D ⊑ B,E ⊑ C are terminological axioms
with respect to a TBox T of a DL knowledge base.
1. If I =
/
0 an audience then
I
T
<
= {hB,Ai,hC,Ai,hD,Bi,hE,Ci} and
χ(I
T
<
) =
(
{hD,B,E,C,Ai};{hE, C, D,B,Ai}
{hD,E,C,B,Ai};{hD, E, B,C,Ai}
{hE,D,C,B,Ai};{hE, D, B,C,Ai}
)
corresponds to the orderings D ≻ B ≻ E ≻ C ≻ A,
E ≻ C ≻ D ≻ B ≻ A, D ≻ E ≻ C ≻ B ≻ A,
D ≻ E ≻ B ≻ C ≻ A, E ≻ D ≻ C ≻ B ≻ A
and E ≻ D ≻ B ≻ C ≻ A.
2. If I = {hB,Ci} an audience then
I
T
<
= {hB,Ci,hB,Ai,hC,Ai,hD,Bi,hE, Ci,
hB,Ei,hD,Ci, hD, Ei} and
χ(I
T
<
) = {hB,Ci,hB,Ai,hC,Ai,hD,Bi,hE,Ci,
hB,Ei,hD,Ci, hD, Ei}
so that χ(I
T
<
) = {I
T
<
}, i.e. χ(I
T
<
) contains ex-
actly one specific audience which corresponds to
the ordering D ≻ B ≻ E ≻ C ≻ A.
ICEIS 2009 - International Conference on Enterprise Information Systems
240

5.2 Determining Consistent Positions
We employ semantic information of arguments as in-
troduced in Sec. 4 and look for answers of concrete
questions, as: does the German team follow one strat-
egy (e.g. to apply an offside trap - cf. Sec. 1.1)?
To use the underlying semantics we determine the
formal structure of the argumentsand apply subsump-
tion reasoning to the argument descriptions (intro-
duced in Sec. 4). For this purpose we use a standard
DL reasoner (Sirin et al., 2007). As a result we get
sub-concepts of the top-level concept Argument (cf.
Fig. 3). Fig. 4 shows arguments and inferred concept
affiliations as well as their defeat relations with ref-
erence to a specific concept-based audience (for sim-
plicity, we have only considered a selection of con-
cepts of our terminology). For arguments shown in
Fig. 4 and a specific audience BZDA ≻ ZDA ≻ MMA ≻
SMA ≻ DPA ≻ ST ≻ A (which is compatible to the
audience I = hZDA,MMAi), a set of arguments
S = {B,E,G} is obtained as a preferred extension.
However, there does not exist any preferred extension
which contains the arguments A and B; hence, claims
about a common strategy among the German team-
mates cannot be supported.
F
E
G
A
B
C
D
{MMA}
{A} {A}{BZDA}
{ZDA} {A}
{MMA}
Figure 4: Arguments (with inferred concept affiliations)
and successful attacks corresponding to a total ordering
BZDA ≻ ZDA ≻ MMA ≻ DPA ≻ ST ≻ A of an audience
I = hZDA,MMAi (see Fig. 3 for abbreviations).
6 CONCLUSIONS
We have presented an approach basically based on
two paradigms: that of argumentation frameworks
and that of description logics. The former is em-
ployed for analysing consistent sets of arguments,
given a set of instantiated arguments at some given
time. The latter is primarily used for defining ter-
minological knowledge in order to characterise argu-
ments at the semantic level; that is to say that instead
of value-based systems, concept-based arguments are
introduced. Concept-based argumentations are more
flexible in those appropriate mappings within an ar-
gumentation system can be automatically adjusted if
we would change the semantic description of argu-
ments. Moreover, we can define preferences among
arguments at the conceptual level instead of taking
simple values, as is the case in value-based argumen-
tation frameworks.
REFERENCES
Amgoud, L., Cayrol, C., Lagasquie, M. C., and Livet, P.
(2008). On bipolarity in argumentation frameworks.
In Int. Journal Of Intell. Systems, 23(10):1062–1093.
Baader, F., Calvanese, D., McGuinness, D. L., Nardi, D.,
and Patel-Schneider, P. F., editors (2003). The De-
script. Logic Handb.: Theo., Impl. and Appl. CUP.
Bench-Capon, T. J. M. (2003). Persuasion in practical ar-
gument using value-based argumentation frameworks.
Journal of Logic and Computation, 13(3):429–448.
Bench-Capon, T. J. M., Doutre, S., and Dunne, P. E. (2007).
Audiences in argumentation frameworks. Artificial In-
telligence, 171(1):42–71.
Bench-Capon, T. J. M. and Dunne, P. E. (2007). Argumen-
tation in artificial intelligence. Artificial Intelligence,
171(10-15):619–641.
Bowers, S., Lin, K., and Lud¨ascher, B. (2004). On integrat-
ing scientific resources through semantic registration.
In In 16th Int. SSDBM, pages 21–23.
Dung, P. M. (1995). On the acceptability of arguments and
its fundamental role in nonmon. reas., logic prog. and
n-pers. games. Artificial Intelligence, 77(2):321–358.
Gruber, T. R. (1995). Towards principles for the design of
ontologies used for knowledge sharing. International
Journal of Human-Computer Studies, 43:907–928.
Patel-Schneider, P. F., Hayes, P., and Horrocks, I. (2004).
Owl web ontology language semantics and abstract
syntax. Technical report, W3C.
Rahwan, I. (2008). Mass argumentation and the semantic
web. Web Semantics: Science, Services and Agents on
the World Wide Web, 6(1):29 – 37.
Rahwan, I. and Banihashemi, B. (2008). Arguments in owl:
A progress report. In COMMA, volume 172 of Fron-
tiers in AI and Appl., pages 297–310. IOS Press.
Sattler, U., Calvanese, D., and Molitor, R. (2003). Relation-
ship with other formalisms. In The Descript. Logic
Handb.: Theo., Impl. and Appl., pages 137–177. CUP.
Sirin, E., Parsia, B., Grau, B. C., Kalyanpur, A., and Katz,
Y. (2007). Pellet: A practical owl-dl reasoner. Journal
of Web Semantics, 5(2):51–53.
Sprado, J. and Gottfried, B. (2008a). Grammar-based argu-
ment construction. In I. Lovrek et al., editor, KES (1),
volume 5177 of LNCS, pages 330–340. Springer.
Sprado, J. and Gottfried, B. (2008b). What motions patterns
tell us about soccer teams. In RoboCup 2008: Robot
Soccer World Cup XII, Suzhou, China. Springer.
Uschold, M. (1998). Knowledge level modelling: concepts
and terminology. The Knowledge Engineering Re-
view, 13(01):5–29.
SEMANTIC ARGUMENTATION IN DYNAMIC ENVIRONMENTS
241