There is a growing interest in holistic ontology engineering approaches that involve multidisciplinary teams consisting of ontology engineers and domain experts. For the latter, who often lack ontology engineering expertise, tools such as Web forms, spreadsheet like templates or semantic wikis have been developed that hide or decrease the complexity of logical axiomatisation in expressive ontology languages. This paper describes a prototype solution for an automatic OWL ontology conversion to articles in Semantic Media Wiki system. Our implemented prototype converts a branch of an ontology rooted at the user defined class into Wiki articles and categories. The final result is defined by a template expressed in the Wiki markup language. We describe tests on two domain ontologies with different characteristics: DMOP and DMRO. The tests show that our solution can be used for fast bootstrapping of Semantic Media Wiki content from OWL files.
There is a growing interest in holistic ontology engineering approaches [
In order to remove the barriers, various tools have been developed that hide or
decrease the complexity of logical axiomatisation in expressive ontology languages.
Among these tools are Web forms [
The aim of this work is to deliver a solution for transformation of OWL files to
Semantic Media Wiki (SMW)2[
The rest of this paper is structured as follows. In Section 2 we discuss the work related to ours. In Section 3, we present a solution for transforming OWL files to Semantic Media Wiki content that is based on SPARQL and user defined Wiki templates. In Section 4 we present a simple evaluation of the implemented solution with two ontologies: DMOP and DMRO. Section 5 contains the discussion, and in Section 6 we conclude. 2
In [
The Halo extension to SMW, contained in SMW+3, was developed in order to facilitate the use of Semantic Wikis for large communities of users and thus consisted of a toolset for increasing the ease of use of SMW features. Among the tools, it provided an import and export functionality for OWL ontologies. Currently, the Halo extension is unmaintained.
MoKi [
The author of [
Our presented solution allows for a direct mapping from an ontology to user defined Wiki templates. In such way, it allows for a direct transformation from arbitrary (user selected) ontology entities to arbitrary Media Wiki form elements (taking into account the distinction between OWL entity types).
The design issues, with which we had to deal, concerned, among others, selecting categories and attributes, and representing many attributes for a single entity. By an attribute we mean a Wiki counterpart of a single OWL property describing a given entity, which is presented mostly in an infobox in our templates.
We can describe hierarchy in OWL files as a set of classes, which can be related with each other by child (subclass), parent (superclass) or equivalent class relation. However, Media Wiki articles can be shown as a directed acyclic graph, where each edge represents child/parent relation, according to its direction. Hence, we decided to treat equivalent (with additional constraints) classes as child classes. Entities described as Named Individuals are candidates for articles. Similarly, classes with subClassOf, equivalentClass or rdf:type attribute are categories stubs. Due to possibility of numerous attributes, we decided to extract only these which are declared by the user in a template file. Wiki markup syntax does not allow to make implicit declaration of a set of values for one attribute. It cannot predict the number of the values for a given attribute without any additional configuration or modification. Hence, we used Semantic Forms4 with parser functions as a solution to this problem. The design choices for the transformation are listed in Table 1. Our approach is described by Algorithm 1. It is necessary to have all mentioned input files in order to define the configuration. To begin the transformation process all classes and individuals which are in a child relation with the root class (including transitivity) have to be found. It can be done by SPARQL queries. We present our solution for classes in Listing 1.1 and for individuals in Listing 1.2. The list of articles stubs is prepared from each acquired entity. Filling all stubs with content from classes and individuals is based on the template and the mapping. In order to do this the ontology has to be searched using SPARQL queries one more time. Article/Category title is based on entity URI. Since there is a requirement for each article to have a title, blank nodes are omitted. The last step consists of transferring the obtained data to Semantic Media Wiki. 4 http://www.mediawiki.org/wiki/Extension:Semantic_Forms Data: OWL les, con guration, template, mapping Result: An updated set of Semantic Media Wiki articles load OWL les; articles process OWL les; for article in articles do if article is a template then
prepare template end if article is a category or an article then set a title to template; fill article with data corresponding to template and mapping; add category footer to article; end end Connect to Semantic Media Wiki; for article in articles do
save article to Semantic Media Wiki; end
Algorithm 1: OWL files conversion to Semantic Media Wiki articles
? s u b c l a s s ( ( owl : e q u i v a l e n t C l a s s / owl : i n t e r s e c t i o n O f / r d f : ,! r e s t / r d f : f i r s t ) j r d f s : s u b C l a s s O f ) + <r o o t C l a s s >. g g
? i n d i v i d u a l r d f : t y p e ? t y p e . ? t y p e ( ( owl : e q u i v a l e n t C l a s s / owl : i n t e r s e c t i o n O f / r d f : r e s t / ,! r d f : f i r s t ) j r d f s : s u b C l a s s O f j r d f : t y p e ) + <r o o t C l a s s >. FILTER ( isURI ( ? i n d i v i d u a l ) && ! isBLANK ( ? i n d i v i d u a l ) ) .
4.1
We prepared an application5 written in Java, which implements the described algorithm. We used Apache Jena6 to read and query OWL files. Tests of the transformation were performed with two different ontologies - DMOP and DMRO.
DMOP is an abbreviation for the Data Mining OPtimization Ontology [
DMRO, a Digital Multimedia Repositories Ontology [
We made simple tests consisting of transforming chosen branches of the ontologies to Semantic Media Wiki.
In case of DMOP, we transformed all classes with related entities, which are in a child relation with DM-Algorithm class (examples in Listings 1.3 and 1.4). The sample Wiki article about C4.5 algorithm (named individual in DMOP) is shown in Figure 2. The article’s title results from the URI of the individual, which is shown in Listing 1.3 (rdf:about attribute) and marked as A in Figure 2). Attributes are labeled as B, C, D, E in the same figure. Due to lack of information in the ontology files not all of infobox values are filled up (they are marked as B and D on Figure 2). Thanks to Semantic Forms, has-quality attribute (marked as F in the figure) has multiple values, what was implicitly declared in the template. Class membership is labeled by G. 5 https://github.com/mimol/owl2wiki 6 https://jena.apache.org 7 http://dublincore.org/documents/dcmi-terms/ 8 http://xmlns.com/foaf/spec/ 9 http://vocab.org/review/terms.html 10 http://obofoundry.org/ro/ 11 http://www.openarchives.org/ore/
Listing 1.3. Structure of a sample DMOP entity classified as an article <owl : NamedIndividual r d f : about=”&PD; C4.5” > <DOLCE L i t e : has q u a l i t y r d f : r e s o u r c e =”&DMOP;
,! H i g h V a r i a n c e P r o f i l e ”/> ( . . . ) </owl : NamedIndividual >
Listing 1.4. Structure of a sample DMOP entity classified as a category <owl : C l a s s r d f : about=”&DMOP; SomeClass”> <r d f s : subClassOf r d f : r e s o u r c e =”&DMOP; SomeSubClass ”/> </owl : Class >
We also successfully conducted another experiment with DRMO whose structure differs from DMOP’s. Although category entities are similarly declared explicitely as OWL classes (as shown in Listing 1.6), the individuals are not declared as named individuals, but are declared as members of owl:Thing (Listing 1.5). Nevertheless, we were able to transform a branch rooted at DMRO:Event.
Listing 1.5. Structure of example DMRO entity classified as an article <owl : Thing r d f : about =”# Event2305”> <r d f : t yp e r d f : r e s o u r c e =”&DMRO Event ; E v e n t S e c t i o n ”/> ( . . . ) </owl : Thing>
Listing 1.6. Structure of example DMRO entity classified as a category <o w l : C l a s s r d f : a b o u t =”&dul ; Event ”>
<r d f s : s u b C l a s s O f r d f : r e s o u r c e =”&dul ; E n t i t y ” /> </ o w l : C l a s s>
We tested our solution with ontologies having different characteristics. Since SPARQL engines by default are not supposed to perform reasoning, SPARQL may turn very structure-sensitive. Although, in the OWL serialization that we used the structure of category entity is based on owl:Class in both, DMOP and in DMRO (Listings 1.4 and 1.3, respectively), there are differences in individual selection. In DMOP, individuals are explicitly declared as a owl:NamedIndividual, while in DMRO they are not. In the latter case, the entities we classify as individuals are instances of owl:Thing. That is why it is important to consider possible cases and take them into account in SPARQL queries or transform OWL files to a canonical representation before SPARQL is applied to query them.
Alternatively, as a more standard solution, we could use an API for handling ontologies like OWL API or Jena ontology API. However, while designing our SPARQL-based solution we kept in mind that it can be further flexibly extended to transform remote (linked) data from SPARQL endpoints to Semantic Media Wiki content.
Our main motivation is the real need for Wiki based tools in the context of such portals as DMO Foundry (http://www.dmo-foundry.org) or OpenML (http: //openml.org). Using the presented in this paper preliminary solution we have generated content that is a human-readable, structured and organised knowledge base. We envisage that it could be used by such domain users as researchers trying to find out which algorithm would match their expectations or even students during classes. 6
In this paper, we have presented a prototype solution for transformation of OWL ontologies to Semantic Media Wiki content. The solution is based on a mapping between selected ontology entities and user-defined Wiki templates, and on using SPARQL. We have successfuly applied our implemented prototype to two different ontologies: DMOP and DMRO.
Despite of describing a working prototype, we still consider this research as a work
in progress. In future work, we plan to consider ideas for cascading templates (similarly
to the work of [