AN ONTOLOGY FOR CERAMICS CATALOGUING

Domenico Cantone, Marianna Nicolosi-Asmundo, Daniele Francesco Santamaria, Francesca Trapani

Department of Mathematics and Computer Science - University of Catania*

Workshop CAA 2015 March 30, April 3 2015 Siena, Italy
*Thanks to Progetto PON PRIME Piattaforma di Reasoning Integrata Multimediale Esperta- Prof. Emiliano Tramontana

OntoCeramic
In the last decades, the need of efficiently organize the classification process of
archaeological finds with particular focus on ceramics has become urgent
for scholars and researchers in the field [6]. Currently, in fact, archaeological
finds cataloguing and classification are mainly performed by traditional methods
like hard-copy archives and standard digital techniques like relational databases.
However, such methods have severe drawbacks. They are tools mainly developed
and maintained in a local way and they usually store partial data, rarely shared
with the whole scientific community, causing an incoherent use of information.
Also, they do not support flexible data-management and information retrieval
algorithms due to lack of advanced reasoning means.
Semantic web is a vision of the World Wide Web in which information carries
an explicit meaning, so it can be automatically processed and integrated by
machines, and data can be accessed and modified at a global level thus allowing
coherence and dissemination of knowledge. Moreover, by means of automated
reasoning procedures, it is possible to extract implicit information present in
data, thus permitting to gain a deeper knowledge of the domain. The definition
of a specific domain is widely called ontology. In the last years, potentiality of
ontologies has been recognized by archaeologists [3, 5]. Some projects have been
undertaken concerning either single typologies of archaeological finds or several
different materials related to each other.
In this contribution we briefly describe our work on OntoCeramic, a Semantic
Web ontology for cataloguing and classifying ceramics. OntoCeramic has been
designed in collaboration between computer scientists and archaeologists, as a first
step to overcome the problem of efficiently mechanize the task of correctly cataloguing
ceramics for the purpose of making such knowledge easily retrievable by scholars and
researchers in the field.
OntoCeramic is an OWL 2 (Ontology Web Language 2) ontology [11] designed
on ICCD (Istituto Centrale per il Catalogo e la Documentazione) data sheets
and on the most important papers in the field [1]. OntoCeramic
consists of 90 OWL classes, 33 object properties, and 20 data properties. It includes
a number of SWRL (Semantic Web Rule Language) rules [12] allowing several
reasoning tasks on the knowledge domain in a short time.
It has been developed using the Protg [7, 2] editor and classified by the
Hermit [8], Pellet [9], and FaCT++ [10] reasoners.
OntoCeramic allows one to carry out many tasks such as associating fragments to a
considered specimen according to its provenance, even indicating from which part of
the vessel they come from, or by find place (i.e., nation, region, or province, and so
on), or by measurements of their parts, or by their colours, or by other features (i.e.,
decoration) [4]. In particular, OntoCeramic allows one to specify shape and type of an
object removing the redundancy of the nomenclature used, reducing ambiguous
classifications of data. In the future we plan to include in the ontology support for
stratigraphic excavations, bibliographic references management including authors and
revisors, and identification of the production factory. We also aim at implementing an
efficient parallelized decision procedure for the language of OntoCeramic that permits
to reason with large data sets.

References
1. Corti L., I beni culturali e la loro catalogazione. Bruno Mondadori, Milano 2003.
2. Gasevic D., Djuric, D., Devedzic V., Model Driven Engineering and Ontology
Development (2nd ed.). Springer. pp. 194. ISBN 978-3-642-00282-3, 2009.
3. La Fragola A., Ontologie distribuite ed interoperabili di supporto allarcheologia,
Draft PhD Proposal, 2005.
4. La Fragola A., LAtlante delle Forme Ceramiche dellEnciclopedia dellArte Antica:
ipotesi di progetto per una gestione e fruizione elettronica dei dati in ambiente
XML, in Bollettino dInformazioni Centro di Ricerche per i Beni Culturali, XII, n.
1, Pisa, 2002, pp. 113-119.
5. Letricot R., Szabados A.V. , Lontologie CIDOC CRM applique aux objets du
patrimoine antique. In Archeologia e Calcolatori, supplemento 5, 2014, pp. 257-272.
6. Moscati P., Archeologia e societ dellinformazione, in
http://www.treccani.it/enciclopedia/archeologia-e-societa-dellinformazione (XXI
Secolo).
7. Protege, http://webprotege.stanford.edu.
8. Shearer R., Motik B., Horrocks I., Hermit: A highly-efficient OWL reasoner. In
Proceedings of the Fifth OWLED Workshop on OWL: Experiences and Directions,
collocated with the 7th International Semantic Web Conference (ISWC2008), Karlsruhe, Germany, October 26-27, 2008, 2008.
9. Sirin E., Parsia B., Cuenca Grau B., Kalyanpur A., Katz Y., Pellet: A practical
OWL-DL reasoner. J. Web Sem., 5(2):51-53, 2007.
10. Tsarkov D., Horrocks I., Fact++ description logic reasoner: System description. In
Automated Reasoning, Third International Joint Conference, IJCAR 2006, Seattle,
WA, USA, August 17-20, 2006, Proceedings, pp. 292-297, 2006.
11. World Wide Web Consortium (W3C). OWL 2 web ontology language
structural specification and functional-style syntax (second edition).
http://www.w3.org/TR/2012/REC-owl2-syntax-20121211/.
12. World Wide Web Consortium (W3C). SWRL: A semantic web rule language.
http://www.w3.org/Submission/SWRL/.

Semantic Web for knowledge management

Semantic Web is a mesh of information linked up in such a way as to be easily processable by machines, on a global scale.
Everything that can be described is an entity (it has an IRI).
Types of Entities:
Class: group of resources with similar characteristics,
Individual: member (instance) of a class,
Object-property: binary relation between two individuals,
Data-property: binary relation between an individual and a value (i.e., numbers).
Ontology: collection of classes, individuals, and properties. It describes the entire knowledge
domain. With ontologies we represent data in such a way that it is understood by computers,
enabling machines to search, aggregate, combine, and share information without a human
operator. Ontologies can automatically extract information from data.
Ontology can be extend with rules ( if-then like expression) which provide additional expressiveness.
The main query language is SPARQL.
Semantic content extraction from unstructured text is an open research field.

OntoCeramic: features and tasks

OntoCeramic aims to cover different aspects of the ceramic classification and cataloguing problem.
To begin with, OntoCeramic helps in identifying unambiguously the location of the finding. One can have many locations to consider for a specific finding such as
province, region, and state. These locations are introduced as OWL classes, and a hierarchy of object-properties is provided in order to specify such locations. The
allowed path is shown in Fig. 1. The figure indicates how locations must be defined for each object. Circles indicate OWL classes, arrows object properties; the
complete hierarchy of these object-properties is shown in Fig. 2 (using the OWL Editor Protg). No other relations (or combinations) are allowed. Properties listed in
Fig. 2 allow one to indicate the next degree of localization. Reasoning tasks concerning finding location are carried out also using the rule set shown in Table 2.2.
OntoCeramic is also able to handle components belonging to a specific object. For example, a cup can be found broken in three disjoint parts that may require distinct
descriptions. Usually, in hard-copy versions of ceramic catalogues, the components of a object are included in a descriptive field of the archive. Information about the
component should be extracted manually by users or by means of research keys. In Ontoceramic, instead, objects as well as fragments are considered as entities. Each
fragment is associated to the object it belongs to by the object property IsFragmentOf. Analogously, each object is associated to its components by the object-property
hasFragment (notice that hasFragment and IsFragmentOf are inverses of each other) and in particular, also by its sub-properties. Sub-properties of hasFragment
are intended to relate an object with one of its fragments taking care of its correct functionality in the object. Thanks to this construction one can precisely describe every
part of an object. Types of fragments that can be implemented are described in Fig. 4.

Fig. 3 shows how properties for fragment management are related to each other. The class of an object is structured in a taxonomy as shown in Fig. 5. To assign a class
to an object, the object-property hasClass is provided, having Object as domain and Class as range. The type of an object is represented in the Object taxonomy.
Sample and sector of a finding are represented by the classes Sample and Sector, respectively. Instances of these classes are associated to an object by means of
the hasSample and hasSector object-properties.

OntoCeramic specifies the color of a finding using the Munsell Color System by means of the data-property hasColor. This property is endowed with three sub-dataproperties, namely hasChroma, hasHue, and hasValue, for Munsell chroma, hue, and value, respectively. In addition, one can provide the discovery date and a
general additional description of the considered object using, respectively, the hasSiteDate and hasGeneralDescription data-properties.
For objects and fragments one can specify its measurements. For example, thickness of an object can be represented by the generic data-property hasThickness and
by its specific sub-properties. The first one, called "hasWallThickness", is used when we indicate the thickness of the object wall, the second one, called
"hasBottomThickness", is used when we indicate the thickness of the foot. The hierarchy of these properties is shown in Fig. 7.
It is possible to specify if a fragment can be physically associated with another fragment to compose a unique object. In this case the "isFittedWith" object-property is
applied. One can indicate the number of the box and the number of the sheet of the hard-copy archive of an object description using a "nonNegativeInteger" value in
"hasBox" and "hasSheet" data-properties, respectively.
To solve a problem concerning the management of the shape of a finding, the "Shape" taxonomy is provided, as shown in Fig. 6. The shape of an object is represented
by the Shape class; instances of this class and of its sub-classes are associated to an object by means hasShape object-property. Every instance of the class
Shape can be uniquely identified by the properties hasFirstShapeDescriptor, hasSecondShapeDescriptor, hasThirdShapeDescriptor, which are sub-properties of
the hasShapeDescriptor data-property. For each data-property a string value can be specified: these values will be the keys for the Shape instances. These
properties can be used as additional human-readable descriptors.
At present, OntoCeramic supports only few types of "shape" and of "shape type", but one can add an arbitrary number of these classes and possibly assert equivalences
among them. Currently, there is no world-wide agreement on the use of a specific nomenclature to indicate the shape and the type of an object. The "Shape" taxonomy
is an attempt to face this problem providing a class for each type of shape and several classes for each specific shape; where required, an equivalence relation can be
established among the shape classes or their sub-classes, to identify shapes which are identical with respect to the classification system but which have been called
with different names. example, "Lamboglia 1A" and "Hayes 8" are represented equivalent. In the future, we aim to provide the complete taxonomy of the Shape class in
order to allow advanced reasoning on it, and to extend OntoCeramic for a complete description of archaeological findings and excavations.