=Paper=
{{Paper
|id=None
|storemode=property
|title=Access and Annotation of Archaeological Corpus via a SemanticWiki
|pdfUrl=https://ceur-ws.org/Vol-632/paper03.pdf
|volume=Vol-632
|dblpUrl=https://dblp.org/rec/conf/semwiki/LeclercqS10
}}
==Access and Annotation of Archaeological Corpus via a SemanticWiki ==
https://ceur-ws.org/Vol-632/paper03.pdf
Access and Annotation of Archaelogical Corpus
via a Semantic Wiki
Éric Leclercq and Marinette Savonnet
University of Burgundy
Le2I Laboratory - UMR 5158
B.P. 47 870, 21078 Dijon Cedex - France
Firstname.Lastname@u-bourgogne.fr
Abstract. Semantic wikis have shown their ability to allow knowledge
management and collaborative authoring. They are particularly appro-
priate for scientific collaboration. This paper details the main concepts
and the architecture of WikiBridge, a semantic wiki, and its application
in the archaelogical domain. Archaeologists primarily have a document-
centric work. Adding meta-information in the form of annotations has
proved to be useful to enhance search. WikiBridge combines models and
ontologies to increase data consistency within the wiki. Moreover, it al-
lows several types of annotations: simple annotations, n-ary relations and
recursive annotations. The consistency of these annotations is checked
synchronously by using structural constraints and or asynchronously by
using domain constraints.
1 Introduction
Document analysis is crucial to archaeologists when trying to understand the
evolution of patrimonial buildings and sites. Documentary sources provide
partial evidences from which researchers will infer possible scenarios on how
a building may have been transformed through the ages. The aim of the CARE
project (Corpus Architecturae Religiosae Europeae - IV–X saec.) is the constitu-
tion of an integrated corpus of the French Christian buildings dated from the
4th to the beginning of the 11th century. It aims at facilitating work of com-
parisons, exchanges and discussions with numerous foreign researchers and
specialists. The project has been launched in France on January 1st, 2008 after
acceptance of the French National Agency for Research and will last 4 years
(2008-2011). More than sixty researchers from about twenty universities, di-
verse research institutions and heritage management institutions are working
on. Various categories of staffs are involved: field archaeologists, historians,
art historians, draftsmen, topographers, PhD students, etc. They are collecting
and analyzing data concerning approximately 2700 monuments. The corpus of
multimedia documents (including texts, maps, and photographies) concerning
every known building will be gradually published in the form of classic books.
The request of a Web 3.0 application with a collaborative component and
the need of document management led us to choose a solution based on a wiki
rather than a database. A prototype is available at http://care.u-bourgogne.
fr. Despite the power of wiki (free input, rich user-interface, traceability, bi-
directional links between pages, etc.), it is difficult to answer a specific query
because of the purely textual information stored. Consequently, an approach
which can provide a semantic annotation of the content is necessary. In addi-
tion, requirements for interoperability and data exchange must be taken into
account since the design phase of the application. The semantic web thereby
provides such kind of solutions by increasing the expressiveness of data repre-
sentation, and by allowing reasoning tools and semantic search.
The computer application part of the project has started in September 2008,
a prototype has been held with MediaWiki and Semantic MediaWiki through
May to July 2009. After this prototyping phase we notice that some functional-
ities are missing in Semantic MediaWiki. For example n-ary relations are not
fully supported and the scope of a tag is generally a document. As in Se-
mantic MediaWiki, annotation can be enhanced as the knowledge evolves. In
most of semantic wiki approaches, subjects of annotation are the whole docu-
ment, we propose a recursive annotation model to cope with different levels
of knowledge granularity as well as extension of domains. In [8], authors pro-
pose an equivalent representation between OWL concepts and Semantic Me-
diaWiki constructs. WikiBridge approach allows to annotate an element with
different annotations in several parts of documents. This functionality can be
used to highlight a specific object described in a document. In [7], authors pro-
vide facilities to ensure the content quality of a wiki, including constraint and
auto-epistemic operators. They introduce semantic checking with three kinds
of constraints that are mostly structural: 1) domain and range; 2) concept cardi-
nality; and 3) property cardinality. In WikiBridge, structural constraints check-
ing is included in the annotation process while domain dependent constraints
are checked asynchronously.
The rest of the paper is organized as follow: Section 2 describes our ar-
chitecture through the physical and logical structure, the semantic layer, the
information access layer. Section 3 concludes the paper.
2 WikiBridge’s architecture
Our proposal is to use MediaWiki to develop a numerical corpus by integration
of individual contributions. We have extended MediaWiki with some DBMS
capabilities and semantic tools: form based acquisition interface, annotations,
query engine.
2.1 Document Structuring
The archaeologists’ work is focused on documents: documentary sources and
documents of excavations are used to analysis of buildings; in result paper
forms are produced. Moreover, document exchange, information retrieval and
Information access
user interactions
Form based annotations Faceted navigation
Web based
Mediawiki Engine
Form based searching
Form based acquisition
(semantic forms)
Summarizing
Persistency
External tools for
loading ontologies
Wiki RDF Ontology
database triple store database
Semantics control
Tag consistency Constraints Verison Control
checking checking
Fig. 1. WikiBridge’s Architecture
integration are uses of these various documents. The multitude of purposes
and the diversity of document content types led to different structuring needs.
Standards such as Open Document Architecture and SGML (Standard General-
ized Markup Language) consider that document has at least two structures of
representation.
The physical structure defines the document presentation. This structure con-
sists of physical elements such as style sheets (CSS, XSLT).
The logical structure defines an organization (relationship of composition, se-
quence) of information contained in the document. This organization repre-
sents the different parts of the document. It is composed of titles, chapters,
paragraphs, notes, figures, etc. Organization of a document in the CARE
corpus is as follows: topography, documentary sources, a succession of
states describing the evolution of architectural building. In each state, plan
of building with concepts of space, architectural elements and function are
known from elements of relative dating such as construction techniques,
building materials, sepulchers etc. This logical structure can not structure
the knowledge and therefore does not allow easy information access.
The logical structure of the document could be stored in a database with
attributes of type LONG, but a specific tool must be developed to display,
to edit the different documents and their structure. Wiki is a suitable tool
for representing these two structures.
The semantic structure has been introduced by other authors [3]. It represents
the information itself i.e. the meaning of document content. The semantic
structure describes information that a user or an agent asks when searching.
It is superimposed on the document and allows to manipulate the rules and
not chapters or paragraphs.
2.2 Physical and logical structure layers
The physical structure is covered by MediaWiki and the logical structure is
managed by Semantic Forms extension1 for MediaWiki. Corresponding mod-
ules are described in light grey in figure 1. Each part of the paper document –
a word file – (figure 2.a) is represented by a model (figure 2.b), models can be
composed. A model is defined by using a mini-scripting language and forms
are created on-the-fly on the basis of models. Two types of acquisition form
have been created: a form for entering a record corresponding to atomic build-
ing and a form corresponding to a group of buildings. Some specific fields
(select lists) and free text based fields are proposed. For instance, they are re-
spectively used for selecting administrative regions of a building and describ-
ing liturgical installations in a building. Finally, a non-expert in archeology
can easily feed the wiki (figure 2.c), by copying and pasting, from paper forms
already made by archaeologists. Results are stored in the wiki database.
Representated by a model
(b) Model: Logical structure
Made by
an archaeologist
(a) Paper document (word file)
Creation of form
Copying and pasting
by a non-expert
(c) Wiki document: Physical structure
Fig. 2. Acquisition form and document model
1
http://www.mediawiki.org/wiki/Extension:Semantic_Forms
2.3 Semantic layer
To improve quality of search, we expanded MediaWiki with semantic compo-
nents (medium grey box in figure 1). Annotations, made by experts, are guar-
anteed by a domain ontology. Experts directly enter and modify annotations
through an extension of the wiki’s editing interface (figure 3) which relies on the
form based annotation component. We restrict access to ontological knowledge
management to a predefined set of Wiki users: we argue that implementing
such functionality without adequate process-level support might have uncon-
trolled consequences on the operation of the overall wiki system. Knowledge
engineers interacting with archaeologists create the domain ontology with stan-
dard tools like Protégé. The scope of domain ontologies includes concepts and
relations of thematic area. Specific extensions of domain ontologies are defined
in the context of a distinct usage of the more general knowledge model [4].
CIDOC Conceptual Reference Model 2 [2] is a domain ontology intended to
facilitate the integration, mediation and interchange of heterogeneous cultural
heritage information. We have made a specific extension of CIDOC ontology
for the European Christian buildings. It consists of:
– found objects : type of buildings, architectural elements (e.g. nave); liturgi-
cal installations (e.g. altar), wall structures and pavements . . .
– religious aspects of these objects: function, consecration;
– spatial aspects: relative position of an object with another;
– architectural evolution of objects: creation, destruction and modification by
adding or deleting element.
A nno ta tion m ade b y
an e xpe rt
Fig. 3. Annotation Interface
2
http://cidoc.ics.forth.gr/
Persistency Two persistency levels have been distinguished:
– A persistence level related to knowledge which includes ontology and an-
notations. We explicitly store in a relational database the conceptual model
defining the structure of the domain ontology (Figure 4). Ontology is loaded
from Protégé by a specific program. As a result, annotations are stored in
RDF data in the RDBMS.
– A persistence level related to document structure is realized by MediaWiki
with the Semantic Forms extension.
has
sub_concept_of
sub_property_of
0..1 1 0..* 0..1 0..*
Concept domain Property has Property_Instance to Instance
0..* 0..* 0..* 1 0..* 0..* 1..*
0..* 0..* 0..* 1..*
range from
0..* 0..* 0,1 0,1
contains
inverse_property
1
Ontology contains
1
Fig. 4. Database schema for ontology
Annotations and consistency checking Two types of annotations have been
identified:
– Simple annotation allows to tag a subject by describing some of its prop-
erties by attributes values (literal) couples. These kind of annotations can
be compared to a restriction on attribute’s domain in the database context.
Theses annotations are mostly related to the ABox level.
– Complex annotation references TBox and ABox levels:
• n-ary relation allows to map a subject with two or more values and
references to other elements (subjects). In this case, some values prop-
erties reference another subject. For example we can annotate an altar
with its dimension, its building material, its location in the nave. The
nave is detailed in another part of the document.
• recursive annotation allows to explain or clarify an attribute by a sub-
annotation which is a simple or a complex annotation.
Moreover, annotations related to the same subject can be expressed in dif-
ferent parts of a document or in different documents. We propose a mechanism
to merge annotations and to visualize all the annotations related to one subject
in the annotation interface.
In order to implement our annotation mechanism, we choose to use the
model of semantic values proposed by Sciore et al. [5] for mediation of rela-
tional databases. They define recursively semantic values by the association of
a context to a simple value. A context is a set of elements which are assignment
of a semantic value to a property. We extend this model by allowing values to
be references to other elements (part of documents, subjects). For the aforemen-
tioned altar example, the annotations are:
1.3(dimension = ”width”, unit = ”m”) (1)
0.95(dimension = ”height”, unit = ”m”) (2)
2.4(dimension = ”length”, unit = ”m”) (3)
marble(buildingM aterial = ”stone”) (4)
#nave143(spatialRelation = ”contained”(spatialP osition = ”center”)) (5)
Annotations (1), (2), (3) should be merged but the semantic values model
treats them as separated annotations. We can introduce an intermediary anno-
tation such as 3D to allow combination of multiples semantic values. The value
is then a specific attribute of the annotation.
dimension = ”3D”(dimensionY = ”width”(unit = ”m”, value = ”1.3”),
dimensionZ = ”height”(unit = ”m”, value = ”0.95”),
dimensionX = ”length”(unit = ”m”, value = ”2.4”))
Annotation modeling using semantic values allows automatic conversion of
units (for example between meters and inches). The same type conversion can
be used for dates from centuries to values interval. Conversion can be used in
query processing or for multi-lingual support.
1.3
value
width
unit m
dimz
0.95
value
#altar23 dimension 3D dimy height
unit m
dimx
2.4
value
length
unit m
Fig. 5. RDF like transformation of semantic values
In WikiBridge, semantic values are reified in a list of atomic annotations
i.e. couples (property, value), related to an object (subject), that are stored as
triples in the database. An identifier is given to each atomic annotation allow-
ing recursive semantic values. Annotation tuples can be translated in RDF and
displayed as graph (figure 5).
Annotations are defined by users through a wizard that controls two kinds
of constraints: 1) Domain values of properties using ABox capabilities; and
2) Structural consistency of properties using TBox capabilities (for instance, a
cathedral can have a nave but cannot have an atrium).
This two kinds of constraints can be checked using the ontology structure
in OWL format. Nevertheless, some domain dependent constraint cannot be
embedded in the structure. For example "a building cannot be dedicated to a
saint before is death date" is represented by the following rule:
isConsecrated(?b,?p) ← hasConstructionDate(?b,?d1) ∧
hasDateDead(?p,?d2) ∧ d1 ≥ d2.
OWL is mainly based on description logics [1] (DL). Some features of DL
make it difficult to use for validating data annotations through integrity con-
straints (IC): 1) OWL-DL works in open world assumption; 2) OWL does not
use the unique name assumption. Finding inconsistent annotations require to
evaluate OWL rules in a closed world assumption to detect violation. The do-
main dependent constraints are checked when users validate an annotation
while domain values and structural properties are checked when users build
the annotation through wizard (figure 3). Three approaches are described in [6]
: 1) skolemisation-based semantics, some constraints are tagged as IC; 2) ruled-
based semantics based on interaction with logic programming that provides
negation as failure under the closed world assumption and 3) query-based se-
mantics that relies on boolean epistemic queries for expressing constraints.
In order to implement constraints two solutions have been tested: 1) transla-
tion of constraints in a programming language such as procedural SQL or PHP
and 2) use of a reasoner and a set of constraints stored in a file. The second
solution was chosen because it allows to define and to add dynamically new
constraints as knowledge evolve.
2.4 Information access layer
Information access layer has been built with taking into account some features
about users. We have thus identified a usage typology in accordance to 1) kind
of usage: reader, investigation, clarification; 2) knowledge degree of the do-
main: domain specialists like archaeologist researchers and non specialists. On
this basis, we can distinguish: 1) general public with a general knowledge of
the area who wants to find information on the known elements; 2) experts un-
derstanding meaning of annotations who need access to detailed information;
and 3) researchers who need to make analysis i.e. cross-checking data from
multiple articles and make emergence of new knowledge.
To handle these different types of users, we offer three types of queries:
1. faceted browsing allows users to explore by filtering available information
through an ontology tree;
2. form based searching provides semantic search by filling in parameters
associated with ontology concepts. Two types of interfaces (figure 6) for
building semantic queries are developed: a wizard lets users to specify
search parameters to engine and users can create query models that are
then stored;
3. aggregate view for each article as factbox.
Fig. 6. Query interface
Three kinds of results can be displayed: 1) results can appear in a list con-
taining links to articles, at the right annotation place, so where the information
is given; 2) user can then manually navigate through articles interlinked; and 3)
users can select annotation to be displayed in the result. From this result, users
can obtain the list of the articles in which have the same annotation. This third
kind of display is a mix of result list and factbox and allows more sophisticated
analysis.
3 Conclusion
A feasible combination of wiki and Semantic Web technologies should pre-
serve the key advantages of both technologies: the simplicity of wiki systems
as shared content authoring tool, and the power of Semantic Web technologies
w.r.t. structuring and retrieving knowledge. In this article, we have demon-
strated that flexibility and data quality required by scientific applications can
be achieved by using wiki with semantic web technologies.
We use annotations to make links between logical layer and semantic layer.
The semantics of annotation is guaranteed by an ontology including constraints
which allows to describe accurately domain knowledge. Our dual approach
allows to cope with evolution of knowledge by modifying the ontology and
annotations dynamically without modifying database schema.
Actually, we only verify structural constraints in a synchronous mode when
users annotate the document. The next version of WikiBridge will automate
verification of integrity constraints by Pellet reasoning engine and annotations
will be marked by an ontology version. Remain the problem of inter-ontologies
version consistency.
Some geomaticians of the Social Sciences and Humanities Research Institute
of Dijon will conduct specific spatial analysis by providing GIS tools from end
of 2010. For thorough analysis, specialized tools (GeoMondrian3 and PostGIS
4
) interconnected by Web Services will be proposed to specifically address the
spatio-temporal aspect. For simple spatial analysis, OpenLayers5 applications
will be developed.
Acknowledgments This work is supported by the ANR (ANR-07-CORP-011).
References
1. Franz Baader, Diego Calvanese, Deborah L. McGuinness, Daniele Nardi, and Peter F.
Patel-Schneider, editors. The Description Logic Handbook: Theory, Implementation, and
Applications. Cambridge University Press, 2003.
2. M. Doerr. The cidoc crm - an ontological approach to semantic interoperability of
metadata. AI Magazine, 24:2003, 2003.
3. Line Poullet, Jean-Marie Pinon, and Sylvie Calabretto. Semantic Structuring Of Doc-
uments. 3rd Basque International Workshop on Information Technology (BIWIT), pages
118–124, 1997.
4. Sebastian Schaffert, Andreas Gruber, and Rupert Westenthaler. A semantic wiki for
collaborative knowledge formation. In In Semantics, 2005.
5. Edward Sciore, Michael Siegel, and Arnon Rosenthal. Using semantic values to facili-
tate interoperability among heterogeneous information systems. ACM Trans. Database
Syst., 19(2):254–290, 1994.
6. Evren Sirin, Michael Smith, and Evan Wallace. Opening, Closing Worlds - On In-
tegrity Constraints. In OWLED, 2008.
7. Denny Vrandecic. Towards automatic content quality checks in semantic wikis. In
Social Semantic Web: Where Web 2.0 Meets Web 3.0, AAAI Spring Symposium 2009,
Stanford, CA, march 2009. Springer.
8. Denny Vrandecic and Markus Krötzsch. Reusing ontological background knowledge
in semantic wikis. In Max Völkel and Sebastian Schaffert, editors, Workshop on Seman-
tic Wikis, volume 206 of CEUR Workshop Proceedings, 2006.
3
http://www.spatialytics.org/projects/geomondrian/
4
http://postgis.refractions.net/
5
http://openlayers.org/