Semantic wikis have shown their ability to allow knowledge management and collaborative authoring. They are particularly appropriate 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 documentcentric 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 allows 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.
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 constitution 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 comparisons, 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, diverse 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, bidirectional 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 addition, 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 representation, 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
functionalities 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
Semantic MediaWiki, annotation can be enhanced as the knowledge evolves. In
most of semantic wiki approaches, subjects of annotation are the whole
document, we propose a recursive annotation model to cope with different levels
of knowledge granularity as well as extension of domains. In [
The rest of the paper is organized as follow: Section 2 describes our architecture through the physical and logical structure, the semantic layer, the information access layer. Section 3 concludes the paper. 2
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
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 s n ed ito s c
a a r ebbW itrse e n u y c n e it s s r e P l o tr n o c s c it n a m e S
Mediawiki Engine
Wiki database Tag consistency checking
Form based annotations Form based acquisition (semantic forms)
RDF triple store Constraints checking
Information access Faceted navigation Form based searching Summarizing
Ontology database Verison Control
External tools for loading ontologies 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 Generalized Markup Language) consider that document has at least two structures of representation.
The physical structure defines the document presentation. This structure consists of physical elements such as style sheets (CSS, XSLT).
The logical structure defines an organization (relationship of composition, sequence) of information contained in the document. This organization represents 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 [
The physical structure is covered by MediaWiki and the logical structure is managed by Semantic Forms extension1 for MediaWiki. Corresponding modules 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 building 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 respectively used for selecting administrative regions of a building and describing 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. (a) Paper document (word file)
1 http://www.mediawiki.org/wiki/Extension:Semantic_Forms 2.3
To improve quality of search, we expanded MediaWiki with semantic
components (medium grey box in figure 1). Annotations, made by experts, are
guaranteed 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
uncontrolled consequences on the operation of the overall wiki system. Knowledge
engineers interacting with archaeologists create the domain ontology with
standard 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 [
Annotation m ade by an expert 2 http://cidoc.ics.forth.gr/ Persistency Two persistency levels have been distinguished: – A persistence level related to knowledge which includes ontology and annotations. 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. sub_concept_of
identified: – Simple annotation allows to tag a subject by describing some of its properties 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 properties 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 subannotation which is a simple or a complex annotation.
Moreover, annotations related to the same subject can be expressed in different 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. [
Annotations (1), (2), (3) should be merged but the semantic values model treats them as separated annotations. We can introduce an intermediary annotation 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.
#altar23 dimension 3D dimy
height dimz dimx width length value unit value unit value unit 1.3 m 0.95 m 2.4 m
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 allowing 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 [
In order to implement constraints two solutions have been tested: 1) translation 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.
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 domain: 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 understanding 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.
Three kinds of results can be displayed: 1) results can appear in a list containing 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
A feasible combination of wiki and Semantic Web technologies should preserve 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 demonstrated 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). 3 http://www.spatialytics.org/projects/geomondrian/ 4 http://postgis.refractions.net/ 5 http://openlayers.org/