=Paper=
{{Paper
|id=None
|storemode=property
|title=Semantics-based Models for the Representation of Claims about Cultural Artifacts and their Sources
|pdfUrl=https://ceur-ws.org/Vol-981/AImWD2013paper2.pdf
|volume=Vol-981
}}
==Semantics-based Models for the Representation of Claims about Cultural Artifacts and their Sources==
https://ceur-ws.org/Vol-981/AImWD2013paper2.pdf
Semantics-based models for the representation of
claims about cultural artifacts and their sources
Aikaterini Vasilopoulou-Spitha1, Antonis Bikakis2,
1
Department of Cultural Technology & Communication, University of the Aegean
2
Department of Information Studies, University College London
1
katvassp@aegean.gr, 2 a.bikakis@ucl.ac.uk
Abstract. Uncertainty and ambiguity are two inherent properties of historical
and archaeological data. It is very often that different researchers express
conflicting opinions about an artifact’s chronology, creation, origin, style or
identification. Ontology-based models, such as CIDOC-CRM, are already
widely used for the representation of cultural artifacts, offering significant
benefits, such as formality, flexibility and extensibility. They were, though,
specifically designed to represent factual cultural data, and not claims or
conflicting opinions expressed about cultural artifacts. In this paper, we
examine how such models may be extended to integrate information about the
sources of cultural information (e.g. bibliographic data) enabling users to assess
the validity of this information. We also propose an alternative approach based
on a more expressive language, N3 Logic. Finally, we discuss argumentation
theory as a tool for the natural representation of claims about cultural artifacts
and the arguments they are associated with.
1. Introduction
The disciplines of both archaeology and history have long used scientific methods,
such as comparison, hypothesis and classification, in order to explain historical
phenomena. And indeed, the explanation of such phenomena, as well as of any others
in the world of our experience, is, according to Hempel and Oppenheim, “one of the
foremost objectives of all rational inquiry and especially scientific research” [16]. But
outside the realm of research and in the field of knowledge representation,
information is often presented as a sum of facts, conclusions or results that the user is
forced to accept as true rather than to inquiry or even doubt about. In this sense for
instance, instead of answering the question “why” an artifact is said to be of a
particular type or style by providing certain evidence, systems present information
that is restricted to answer the question “what” has been said about this artifact: in this
case that it is of a particular type or style.
In view of the need for more expressive and explanatory forms of information
presentation, capable of making available, apart from metadata related to a specific
artifact in the form of mere statements, also the resources that assert or justify them,
the current paper proposes the use of bibliographic data (such as books) as clues for
supporting information regarding an artifact’s date, location, style or identification.
Following the Hempel-Oppenheim theory, in which explanation is divided into two
sections: the explanans, namely the statements from which the conclusion (or
explanandum) can be deduced; and the explanandum, which is the logical
consequence of the explanans, we attempt to explain a “phenomenon”, here a specific
artifact perceived as a sum of cultural and historical features, by means of an
explanandum, a conclusion in the form of a statement, and an explanans, a premise in
the form of a resource (e.g. a book), which is based on a certain claim or argument (i.e
a scientific opinion) made about this statement.
We use as an example the case of the sculpture “Ephebe of Marathon”, a Greek
bronze statue, which is conserved in the National Archaeological Museum of Athens,
and a statement on the style of the sculpture: “Ephebe of Marathon follows the
Praxitelian style” (the explanandum). This statement is derived from Kaltsas’ book
(the explanans) entitled “The National Archaeological Museum” where he claims:
“The Praxitelian style is manifest in this artwork, that is why is regarded by many
researchers as the work of the school of one of the most renown Greek artists of the
4th century BC, Praxiteles (around 340-330 BC)” [17]. We argue that this type of
information management, which employs bibliographic resources as data sources
referring to cultural metadata, allows the user to assert the provenance and validity of
the information provided as well as it enables the development of a whole new
network of associations between cultural artifacts; in this case between the Ephebe of
Marathon and other “praxitelian” artworks, such as the Hermes of Praxiteles, a statue
also verified as “praxitelian” by the Greek geographer and traveler Pausanias, in his
work entitled “Description of Greece” (5:17:3)1.
Considering the significance of making such information accessible and retrievable
for a researcher, a student or even for a common user, the present paper intends to
discuss and propose potential solutions for the challenges encountered in the process
of representing claims about statements by linking these different types of data:
cultural and bibliographic. First, we analyze the capabilities and limitations of
CIDOC-CRM [12], an ontology written in RDF that has been established as standard
for the representation of cultural information, and propose minimal extensions that
enable linking statements about cultural objects with the relevant bibliographic
information (e.g. the bibliographic resource, and its author(s)). Second, we present a
logic-based approach that enables a more natural integration of different types of
information and overcomes expressive limitations of RDF. And third, we discuss how
AI tools, such as argumentation theory, may additionally enable the representation of
claims made about cultural objects and the arguments that led to those claims.
The remainder of the paper is structured as follows: Section 2 presents background
information on semantics-based representation models for cultural and bibliographic
information. Section 3 describes potential semantics-based solutions for the problem
of integrating these two types of information. Section 4 discusses the potential use of
argumentation theory as a tool for the representation of claims about cultural artifacts
and the reasoning behind the claims, and Section 5 concludes.
1
http://www.perseus.tufts.edu/hopper/text?doc=Perseus:abo:tlg,0525,001:5:17:3
2
2. Background
The vast growth in the volume of digitized cultural data along with the current need of
cultural heritage institutions, such as museums, libraries and archives, to interlink
their datasets in large scale has raised the demand for more complex forms of data
representation. In this process, pivotal is the role of semantic technologies, such as
ontologies, as the examples of the Amsterdam Museum2 and the Europeana3 portal
clearly manifest. Following these developments, the semantic models used for
encoding cultural data, whether they are museum, bibliographic or archival, have
been designed to correspond to this new need for mergence. Regarding museum
information, CIDOC-CRM (Conceptual Reference Model) is one representative
example of how such a model can map and integrate data coming from different
sources in a global, extensive, and machine-readable way using the RDF semantics.
Its event centric mechanism that employs a broad vocabulary (presently counting 86
classes and 137 unique properties) enables the inter-relation between people, things,
places and time-spans through common events. On the other hand, the shift of
attention in data integration has enforced the creation of more hybrid data models.
One characteristic example of the new data representation strategies that are being
currently yielded is FRBRoo4, an entity relationship model comprised of four main
entities: “work”, “expression”, “manifestation”, and “item” according to the IFLA
FRBR standards which aims to serve as an integrated ontology for both museum and
bibliographic data by combining FRBR and CIDOC CRM classes and properties.
The idea of linking cultural and bibliographic data, as expressed in FRBRoo, has
been the result of the huge discussion, initiated recently, on the potential of modeling
not only the metadata about a cultural artifact in the form of concrete statements but
also the resources (i.e bibliographic) that verify them, as it has also been envisioned
by Le Boeuf [9]. In fact, the ability to assert and assess the provenance of cultural
metadata is closely related to the interdisciplinary nature of humanistic research itself,
at the core of which stands the process of comparing, referring and verifying the
assembled information resources [3].
3. Problem Analysis
In the context of investigating and experimenting with more complex forms of data
documentation able to commonly conceptualize both cultural and bibliographic
information, the present analysis brings for discussion an issue that requires further
research. This refers to the process of modeling claims concerning statements about
artifacts, as they appear in bibliographic resources. In this sense, main objective of the
current study is to semantically represent the integration of cultural and bibliographic
information, such as Kaltsas’ opinion about the Ephebe of Marathon, which is
recorded in his book “The National Archaeological Museum” and allows the
2
http://semanticweb.cs.vu.nl/lod/am/
3
http://www.europeana.eu/portal/
4
http://www.cidoc-crm.org/docs/frbr_oo/frbr_docs/FRBRoo_V1.0.2.pdf
3
connection with the Hermes of Praxiteles in terms of style according to Pausanias’
claim in his “Description of Greece” (5:17:3).
3. 1 Modeling claims in CIDOC CRM: barriers and limitations
Semantic models for cultural heritage, such as CIDOC CRM, aim to conceptualize
and map data as concrete statements in the form of “Ephebe of Marathon has
praxitelian style” rather than as statements that result from certain claims, such as
“The National Archaeological Museum states that Ephebe of Marathon has
praxitelian style”. In RDF terms, the last part of this sentence (i.e Ephebe of Marathon
has praxitelian style) is not a mere class but a whole new triple. In order to represent
the source of the above claim too, that is the bibliographic resource, the whole
sentence in RDF would be formed as follows:
[Ephebe of Marathon has praxitelian style] (Subject)
[is referred to by] (Predicate) (1)
[“The National Archaeological Museum”] (Object).
Using terms of the CIDOC CRM vocabulary, the subject of the above statement
would be a RDF triple consisting of an instance of the E24 Physical Man-Made Thing
class (Ephebe of Marathon) as the subject, an instance of the E55 Type class
(Praxitelian style) as the object, and the P2 has type property. The P67 is referred to
by property and an instance of the E22 Man-Made Object class (“The National
Archaeological Museum”) would then be the predicate and object of statement (1),
respectively. Information about the author of the book could be provided through the
path: “The National Archaeological Museum” (E22), P108B was produced by, an
appropriate instance of the E12 Production class, P14 was curried out by, “Nikolaos
Kaltsas” (instance of E39 Actor). The P14.1 in the role of property and an instance of
E55 Type class (author) are further used to specify Nikolaos Kaltsas’ role as the
author of this production (Figure 1).
Hence, in this case, the carrier of the particular claim (Nikolaos Kaltsas) is not
represented directly as the object of statement (1), but indirectly as the creator of its
bibliographic source due to a limitation of the underlying model: by linking the claim
directly to its carrier, and the carrier to the source of the claim, we would miss the
connection between the claim and its source, as more than one productions (e.g.
books) may have been carried out by the same actor (e.g. author).
3.2 RDF reification: challenges and barriers
One way of solving this problem is to employ a mechanism which is enabled by the
RDF language and is based on the use of rdf:statement, called “reification”.
rdf:statement
is the class of RDF statements, and the domain of three properties:
rdf:subject, rdf:predicate
and rdf:object,
which associate a statement to its parts.
Here, we could define the subject of statement (1) as an instance of rdf:statement,
and use these three properties to link this statement with its subject (Ephebe of
4
Marathon), its predicate (P2 has type), and its object (Praxitelian style). Reified
statements are not currently supported by CIDOC CRM. They are, though, part of the
standard RDF semantics. Subsequently, P67 is referred to by property may be used to
connect the subject of statement (1) with the “The National Archaeological Museum”.
the claim
E24 Physical Man-Made Thing
“Ephebe of Marathon”
P2 has type the source of the claim (bibliographic resource)
E55 Type E22 Man-Made Object
P67 is referred to by
“praxitelian” “The National Archaeological
Museum”
P108B was produced by
the person who made that claim
E39 Actor P14 was carried out by
E12 Production
“Nikolaos Kaltsas”
P14.1 in the role of E55 Type
“author”
Figure 1: Representation of Kaltsas’ claim and its bibliographic resource in CIDOC CRM.
On the other hand, setting an rdf:statement
as the domain of a property, here P67,
would actually mean diminishing the ability of CRM for self-containment, since such
a class is not a component of the CIDOC semantics. From this aspect, only a CIDOC
CRM statement (a crm:statement) could conform with CRM standards, which due to
a current lack of specification in CIDOC CRM needs to be appropriately defined.
Similarly to the properties attached to rdf:statement
(rdf:subject, rdf:predicate,
rdf:object), we define crm:statement as the class of all statements containing terms
from CIDOC CRM. crm:statement is the domain of three properties: crm:subject,
crm:predicate
and
crm:object. The range of crm:subject and crm:object is E1 CRM
Entity, since E1 is the top entity of the entire conceptual model, and therefore can be
related to any class - given that all the classes are its subclasses. crm:predicate
is then
used to refer to any possible relation between instances of E1 – in other words to any
relation between the subject and the object of a crm:statement. This type of property
serves as the superproperty of all properties of CIDOC-CRM and hence allows the
linkage between any two of its classes. According to CRM rules and naming
conventions, in which each property is represented by character “P” and a number
(property id), as well as a name, we name this property as P0 CRM Property. The
domain and range of P0 is E1 CRM Entity. This enables P0 to link any individuals of
the CRM model. We also define P0 as the range of crm:predicate in order to
represent any relation that may connect any two individuals in a CIDOC CRM
5
statement (crm:statement). Figure 2 shows how a
crm:statement, as
specified above,
may be used to represent our example.
is referred to by
crm:statement CRM Class
crm:subject crm:predicate crm:object
E1 CRM Entity P0 CRM Property E1 CRM Entity
E22 Man-Made Object
E24 Physical Man-Made
E55 Type
Thing
“The National
P2 has type
“praxitelian” Archaeological Museum”
“Ephebe of Marathon”
Figure 2: crm:statement,
its properties and instances describing the running example.
The ability of P0 to link any class of the CRM model and hence express a huge
volume of relationships is actually enabled by the specification of E1 CRM Entity as
its range and domain. Aside the property id, name, range and domain, other issues,
such as quantification and scope note, need also to be considered. Quantification
refers to the so called “property quantifiers”, a term used for the “declaration of the
allowed number of instances of a certain property that can refer to a particular
instance of the range class or the domain class of that property” [12]. In this case, the
quantification is unconstrained. This practically means that E1 CRM Entity, which is
the range and domain class of P0 CRM Property, may have zero, one or more
instances of P0 (many to many (0,n:0,n)). Scope note, on the other hand, is the textual
description of the meaning of the class or property. In this sense, P0 CRM Property
can be described as the property that can take the meaning of any property in CIDOC
CRM. Figure 3 depicts all these specifications along with an example, which is based
on our running example. In this case, P0 CRM Property is superclass of the P2 has
type property, which links E24 to E55.
P0 CRM Property
Domain: E1 CRM Entity
Range: E1 CRM Entity
Quantification: many to one (0,n:0,n)
Scope note: This property can take the form of any property in CIDOC CRM
Example: Ephebe of Marathon (instance of E24) has type (P2 sub property of CRM
Property P0) Praxitelian style (instance of E55)
Figure 3: P0 CRM Property specification
6
Notwithstanding its apparent advantages, reification is a mechanism that has been
heavily criticized by the semantic community, among others also by Tim Berners-Lee
[5], who claims that “the form of reification which is provided by the original RDF
specification is not suitable”. McDermott and Dou, in particular, argue that the
creation of additional triples as a means of expressing statements results in a blow up
of size due the increased number of the triples used [19]. The major issue, however, is
that “reification triple is unrelated to the reified triple in the knowledge base” [23],
since “if we choose to assert each triple as well as its reification, then it is asserted
unconditionally” [21]. Consequently, “RDF reification does not assert the original
triple”, which means that the CIDOC-CRM triple comprised of an instance of class
E24, the P2 has type property and an instance of class E55 cannot actually relate to
the above reified triple (crm:statement). As a result, such a statement is “rather
cumbersome to query with SPARQL language” [15].
All these reasons explain why reification has been confronted with so much
skepticism. Yet, it is also worth mentioning that it does provide a solution to the
inability of RDF triples to express more complex datasets without requiring any
further changes in the semantic structure of CIDOC-CRM, since reification is the only
tool for modeling statements within RDF, the semantic language upon which CIDOC-
CRM is based. However, the problematic nature of reification itself forces us to seek
alternative ways of solving the present problem.
3.3 A rule-based approach: the use of N3 formulae
In view of the above need, the current study proposes the use of rules and rule-based
reasoning. Rules are, according to [2], a key for the realization of more advanced
reasoning capabilities for web applications. Furthermore, the implementation of logic
and rules facilitates data integration, which is important when dealing with
heterogeneous data, like in this case, since it allows us to make inferences which, in
turn, enable data translations. Since RDF structure does not include rules a rule
language could be used as its extension. One candidate “rule language” is N3 Logic,
which aims to make a minimal extension to the RDF data model so that the same
language could be used for logic and data” [6]. Given that CIDOC CRM is structured
according to the RDF semantics, this solution could serve the purposes of the current
study.
With respect to the current problem, N3Logic could provide the means of
addressing statements by extending RDF triples with features, such as quoted
formulae. According to the N3 syntax, a formula allows statements to be made about,
and to query, other statements by grouping triples to sets, a process that is often
referred to as “quoting” [7]. “Quoted formulae allow N3 formulae to be quoted within
N3 formulae” using braces “{…}” [6]. It is due to “quoting” that it is possible to
check the provenance of information, namely to distinguish who states (or believes)
what. Another advantage of using quoted N3 formulae is that statements about other
statements can be described within the same formula. This is, in fact, what
differentiates a quoted formula from an RDF triple, since in the RDF language the
ability to address a statement to another statement requires the addition of another
triple (reified triple) along with the original one, as already discussed above.
7
In the context of the present analysis, we use N3 formulae to represent Kaltsas’ and
Pausanias’ beliefs, as they are expressed in their books, and allow the stylistic
connection between the two artifacts (Hermes of Praxiteles and Ephebe of Marathon).
In this sense, “Description
of
Greece
states” is a phrase that indicates the source of
information, while the N3 formula “{mit:
Hermes
of Praxiteles
has
type
style:praxitelian}.” represents the information itself. The variable “style” is also added
to form the statement. Ultimately, the whole formula can be read as: “Description of
Greece states that Hermes of Praxiteles has praxitelian style”. Similarly, information
about the Ephebe of Marathon can be represented in an N3 formula where the phrase
“The
National
Archaeological
Museum
states”
indicates who has made the statement,
whereas the formula {mit:
Ephebe
of
Marathon has
type
style:
praxitelian}
expresses
the content of the statement itself. Thus, one can read “The National Archaeological
Museum states that the Ephebe of Marathon has praxitelian style”. Hence, from these
two N3 formulae one concludes, that both the “Hermes of Praxiteles” and the
“Ephebe of Marathon” have praxitelian style, which should be considered to be true.
In order to indicate the source of Kaltsas’ claim which is a bibliographic resource (a
book), we can use the following N3 formulae:
The
National
Archaeological
Museum
states
{mit:
Ephebe
of
Marathon has
type
style:
praxitelian}.
(2)
{The
National
Archaeological
Museum
is
type:
book}.
{Kaltsas
is
type:
author
of
The
National
Archaeological
Museum}.
The same applies regarding Pausanias’ claim, as it is recorded in his book. In
particular, the N3 formula “{“The
National
Archaeological
Museum”
is
type:
book}.”
represents the information that “The National Archaeological Museum is a book”,
whereas the N3 formula“{Kaltsas
is
type:
author
of
“The
National
Archaeological
Museum”}.” represents the information about the role of Kaltsas as the author of the
book “The National Archaeological Museum”. In the first formula, the use of the
keyword “is” and the variable “type” is appropriate, while in the second one the
addition of the keyword “of” is used to link the author Kaltsas to the book “The
National Archaeological Museum”.
Hence, the ability to “quote” claims, provided here by N3Logic, can be a very
expressive tool for addressing statements made about other statements which in the
current study enables the interwieving of arguments (i.e “Ephebe of Marathon has
praxitelian style” and “Hermes of Praxiteles has praxitelian style”) whose origin can
be known (Kaltsas – Pausanias) and assessable (“The National Archaeological
Museum” -“Description of Greece” (5:17:3)). Furthermore, as Berners- Lee argues in
[6], it “allows rules to integrate smoothly with RDF”, making N3 an important
candidate in the course of extending RDF. In addition, the combination of ontologies,
such as RDF, with rules, such as N3 rule language, as it was attempted here, is a
practice that conforms to current W3C standards [14]. Nevertheless, it is a nontrivial
issue, since different structures make the overcoming of the impedance mismatch
quite hard [8]. At this stage, the present research concentrates more on investigating
N3 ability to address statements made about statements as alternative to RDF
reification and less on the handling of these obstacles. Concluding, notwithstanding
8
N3 successful way of solving the current problem through “quoting”, its rather simple
“vocabulary” leaves the question of whether it is able to describe the complex
structure of arguments and their sources open for further consideration.
4. Open Issues and Future Directions
Linking statements about cultural objects to the people who made them, or to the
bibliographic resources that these statements are derived from, and making this
information publicly available, is only the first step towards explaining why these
statements were made, and helping people to assess their validity. For example, a
claim made in a scientific journal by an expert is always considered more valid than
one made by a journalist in a local newspaper. However, even experts may sometimes
disagree and express conflicting claims about a cultural object.
Such is the case of our current example and the contradictory theories that have
been built around the question whether Hermes of Praxiteles is actually an original
artwork of the Greek artist Praxiteles or a roman copy. In the context of this
discussion, Oscar Antonsson claims that it is Praxiteles’ original artwork, yet it has
endured some substantial alterations implemented by a roman sculptor later on.
According to this theory, the original artwork initially represented Panas and not
Hermes due to the existence of some marks at the back of the statue that signify the
existence of some kind of cloth, probably of animal skin, pointing to Panas. On the
other hand, the archaeologist Dora Katsonopoulou, in one of her articles about
Hermes of Praxiteles, criticizes Antonsson’s theory arguing that though attractive, his
argument about Panas is rather weak, since the cloth could be used for Hermes. Her
counter-argument is based on the tradition according to which it was Hermes and not
Panas that carried his brother Dionysos to the Nymphs, as depicted in the complex.
For cases like this, the provenance of the claims may not be sufficient to assess
their validity. One may also need to examine and analyze the reasoning that led to
these claims, and apply certain criteria to resolve any potential conflicts. Below, we
discuss these issues in more detail and propose potential solutions from the field of
Artificial Intelligence.
4.1 Exposing the reasoning behind claims
As we already argued in the introduction, ontology-based models, such as CIDOC-
CRM aim at answering questions about cultural objects of the form: “What is known
about a certain object?” or “What has been said and written about a certain object?”.
On the other hand, the extensions that we propose in section 3 give answers for
questions of the form: “Who made a certain claim about a certain object?”. As we
argue above, one may also want answers to the question: “How was a certain claim
made?” or, in other words, “what is the reasoning behind a certain claim?” in order
to be able to assess its validity. Let’s take for example Antonsson’s claim that Hermes
of Praxiteles initially represented Panas and not Hermes. Assessing this claim may be
based on assessing Antonsson’s expertise as an art historian, but also on evaluating
the argument that led to the particular claim: that there are marks that signify the
9
existence of some cloth from animal skin, which point to Panas. We, therefore, need a
way to represent this argument and link it to the claim that is supports.
Argumentation theory is a tool from Artificial Intelligence, which studies the
formation of arguments for or against a certain claim, as well as the relations between
these arguments. For example, Dung’s abstract argumentation framework [13] is a
tuple , where A is a finite set of arguments and → is a binary attack relation on
A×A. Other frameworks, such as the bipolar AFs [11] also define a support relation
among arguments. Using such frameworks, we will be able to represent not just the
final claim, but also the reasoning behind or against the claim, by defining the
arguments that support or respectively oppose the claim. In our example, Antonsson’s
argument (A), and Katsonopoulou’s counter-argument (K), according to which the
cloth could be used for Hermes, since the tradition recognizes him as the person who
carried Dionysos to the Nymphs, may be graphically represented as shown in Figure
4, where the arrow denotes an attack from B to A.
Figure 4: Antonsson’s and Katsonopoulou’s arguments in Dung’s framework.
4.2 Resolving potential conflicts
When all the available arguments for or against claim are made available, one may be
able to judge which if these arguments are valid, and reach a conclusion about the
validity of the final claim. However, in more complex cases, where the number of
arguments is big, and the arguments are interrelated in many different ways, it may
not be straightforward to reach a conclusion just by looking at the available
arguments. In such cases, users may want the system to be able to draw these
conclusions on their behalf, possibly using their feedback.
Argumentation Frameworks may offer solutions to this problem too. AFs are not
just representation models. They also define acceptability semantics, namely the set of
arguments that are accepted as valid, given the overall set of arguments and their
relations. For example, in Dung’s AF, roughly an argument is accepted if all the
arguments attacking it are rejected; and it is rejected if it has at least an argument
attacking it which is accepted. The set of accepted arguments in this framework is a
set of arguments that does not contain any argument attacking another argument in the
set. In our example, according to Dung’s acceptability semantics, argument K
(Katsonopoulou’s argument) will be labeled as accepted, since it is not attacked by
any other argument, and argument A (Antonsson’s argument) will be rejected, since it
is attacked by an accepted argument (A). Preference-based argumentation
frameworks, e.g. those described in [1,4,20] extend Dung’s framework with
preference criteria, e.g. related to the structure of arguments or to their provenance, or
even user-defined criteria, to determine whether an attack is successful or not, i.e.
whether it is sufficient to invalidate the argument under attack. In our example,
preference criteria that can be used to determine whether K successfully attacks A,
10
may include the expertise of the people who made the arguments (Antonsson,
Katsonopoulou), their gender, nationality and education [22], since, for instance,
nationality may affect the impartiality of an argument. Other parameters concern the
chronology of the arguments (e.g. recent arguments may be considered stronger than
older ones as they are based on new evidence) and the validity of the evidence that
supports each argument. It is among our future plans to define a general set of
preference criteria that may be applied for the comparison of arguments about cultural
artifacts.
4.3 Linking statements, arguments and decisions
Applying argumentation theory enables formally representing arguments for or
against certain claims, and evaluating the acceptable arguments by applying user-
defined preference criteria. A further issue, which naturally arises, is how to represent
these different types of information (statement, bibliographic references, arguments)
in a common model, which is readable by human users, but may also be processed by
machines. A common representation model will facilitate the management and
processing of the available data. It will also enable creating views that cover all
different aspects of cultural information.
The SIOC Argumentation Vocabulary [18] will enable us to create the link
between all available information elements: the statements about the cultural objects,
the bibliographic references that these statements are derived by, the arguments that
are made for or against certain claims, and the set of acceptable arguments. The
vocabulary includes concepts for the representation of statements, arguments about
statements, support and attack relations between arguments (these are available in the
extension proposed in [10]) and decisions on the acceptable arguments. By applying
appropriate SPARQL queries, one may then retrieve more precise information about
certain statements and arguments, or aggregate information e.g. about the set of
statements that are evaluated as valid.
5. Conclusion
The semantic representation of claims regarding cultural metadata (i.e an artifact’s
style) along with their origin (i.e the person who made them) and their source (i.e
their bibliographic expression), as it was attempted in this paper, is a step towards the
realization of more sophisticated user queries in systems of higher intelligence able to
process and further assess the provenance and validity of the arguments and the
supporting evidences (i.e historical, philological or archaeological) upon which these
claims are structured, prioritizing certain ones against others. Modeling arguments in
the field of historical and archaeological research by employing Argumentation
Frameworks, such as Dung’s AF, and argumentation vocabularies, such as SIOC, is a
challenging task that entails many difficulties due to the diverse nature of cultural data
(e.g vagueness, uncertainty, ambiguity) that these arguments are referring to and the
present study intends to investigate in the future.
11
References
[1]. Amgoud, L. and Cayrol, C.: A Reasoning Model Based on the Production of Acceptable
Arguments. Annals of Mathematic and Artificial Intelligence, 34(1-3) (2002) 197-215
[2]. Antoniou, G.,Wagner, G.: Rules and Defeasible Reasoning on the Semantic Web, RuleML
2003, LNCS 2876. (2003) 111-120
[3]. Benardou, A.: Classical Studies facing Digital Research Infrastructures: From Practice to
Requirements. In Digital Classicist & Institute of Classical Studies Seminar. (2011)
[4]. Bench-Capon, T.J.M.: Persuasion in Practical Argument Using Value-based
Argumentation Frameworks. Journal of Logic and Computation, 13(3) (2003) 429-448
[5]. Berners-Lee, T.: Reifying RDF (properly), and N3 [online]. Available at:
http://www.w3.org/DesignIssues/Reify.html (2004)
[6]. Berners-Lee, T., Connolly, D., Kagal, L., Scharf, Y.: N3Logic: A Logical Framework For
the World Wide Web. Journal Theory and Practice of Logic Programming 8 (3) (2008)
249-269
[7]. Bizer, C.: Quality-Driven Information Filtering in the Context of Web-Based Information
Systems. In Fachbereich Mathematik und Informatik Arbeitsgruppe Netzbasierte
Informationssysteme. Freie Universität Berlin (2007)
[8]. Bruijn, J. d., Eiter, T., Tompits, H.: Embedding Approaches to Combining Rules and
Ontologies into Autoepistemic Logic. In Proc. Eleventh International Conference on
Principles of Knowledge Representation and Reasoning, AAAI Press (2008) 485-495
[9]. Boeuf, P. L. : Using an ontology-driven system to integrate museum information and
library information. In Symposium on Digital Semantic Content across Cultures. Paris, the
Louvre (2006)
[10]. Cabrio, E., Villata, S., and Gandon, F.: A Support Framework for Argumentative
Discussions Management in the Web. In Proc. ESWC-2013 (accepted)
[11]. Cayrol, C. and Lagasquie-Schiex, M.C.: Bipolarity in argumentation graphs: Towards a
better understanding. In Proceedings of SUM, LNCS 6929 (2011) 137-148
[12]. Doerr, M.: The CIDOC Conceptual Reference Module: An Ontological Approach to
Semantic Interoperability of Metadata. AI Magazine 24(3) (2003) 75-92
[13]. Dung, P.: On the acceptability of arguments and its fundamental role in nonmonotonic
reasoning, logic programming and n-person games. Artificial Intelligence, 77(2) (1995)
321-358
[14]. Eiter, T., Ianni, G., Krennwallner, T., Polleres, A.: Rules and Ontologies for the Semantic
Web. In Reasoning Web 2008, LNCS 5224 (2008) 1-58
[15]. Heath, T., Bizer, C.: Linked Data: Evolving the Web into a Global Data Space. Synthesis
Lectures on the Semantic Web, Morgan & Claypool Publishers (2011)
[16]. Hempel, C. G., Oppenheim, P.: Studies in the Logic of Explanation. Philosophy of
Science, 15 (2) (1948) 135-175
[17]. Kaltsas, N.: The National Archaeological Museum. Foundation of Ioannis S. Latsis (2008)
[18]. Lange, C., Bojars, U., Groza, Breslin, T.J. and Handschuh, S.: Expressing argumentative
discussions in social media sites. In Proc. of SDoW (2008)
[19]. McDermott, D., Dou, D.: Representing Disjunction and Quantifiers in RDF. ISWC 2002,
LNCS 2342 (2002) 250-263
[20]. Modgil, S.: Reasoning about preferences in argumentation frameworks. Artificial
Intelligence, 173(9-10) (2009) 901-934
[21]. Watkins, E. R., Nicole, D. A.: Named Graphs as a Mechanism for Reasoning About
Provenance. APWeb 2006, LNCS 3841 (2006) 943-948
[22]. Wyner, A., Scheider, J.: Arguing from a point of view. In First International Conference
on Agreement Technologies. Dubrovnik, Croatia, October 15-16 (2009)
[23]. Yrjölä, J.: Trusting the content of web resources in the Semantic Web. In TKK T-110.5290
Seminar on Network Security (2007)
12