In recent years a lot of e ort has been put into research dealing with accessing relational data with the help of ontologies. Solutions related to this issue focus on the Ontology-based Data Access (ODBA) architecture. Therefore the family of DL-Lite formalisms and in particular DL-LiteA has been developed to enable ontology design for e cient processing of conjunctive queries targeting a large amount of data stored in relational data bases. However, due to strong limitations and the comparatively high complexity of constructs comprised by DL-LiteA, it can be quite a challenging task to design ontologies which are based on this formalism. This paper presents an editor for DL-LiteA ful lling ontology design. Thanks to its features, such as automated generation of axioms, on-the- y inferring and graphical representation of relations between concepts, the editor simpli es the design process and improves the quality of the resulting ontologies signi cantly. The application and evaluation of the editor in two projects on data and knowledge integration for energy saving and urban planning con rmed these advantages.
Visualization and GUI based editing of formal ontologies have been central to
various projects since the introduction of the term ontology in the engineering
and technological disciplines by Gruber [
At rst, when designing an ontology which has to follow the syntax declared in one of the OWL pro les or in one of the DL logics like AL, SHOIN ,SHIQ, E L etc., the programming environment (i.e. ontology editor) should support the designer by taking these syntactical restrictions into account. The designer should be noti ed if a speci cation of an ontology element is not aligning with the de nition of the selected pro le or logic; for example if such speci cation uses a construct that is not available in the selected logic. Alternatively the environment should support only those constructs that are de ned within the selected pro le or logic, i.e. the GUI available for a designer must not contain constructs which aren't de ned.
However, the grammar control of the ontology designer is only one challenging aspect. The second task concerns on-the- y validation of semantics. Though, this issue appears to be more di cult to address since it requires on-the- y reasoning, the degree of complexity and response time depends on the properties of the selected logic and the ontology being edited. In spite of the expected di culties concerning the realization of these features, this demand seems to be just natural. Analogous to that in the area of software engineering, on-the- y code validation belongs to the state of the art in this eld and is implemented in numerous IDEs, such as eclipse1.
In this paper we present the ontology editor ClickOnA, that implements an approach addressing two challenges mentioned above, i.e. taking into account the constructors and syntax available in a particular description logic (to be precise it is DL-LiteA) during the execution design/editing operations. The DLLite family and in particular DL-LiteA has been developed as a fundamental element of the Ontology-based Data Access (ODBA) architecture to facilitate ontology design for e cient processing of conjunctive queries targeting a large amount of data stored in relational data bases.
ClickOnA has been applied for the project SEMANCO2 (Semantic Tools for Carbon Reduction in Urban Planning) focusing on implementation of analytic and data management services for carbon reduction in urban planning. Due to its features, such as automated generation of axioms, on-the- y inferring and graphical representation of relations between concepts, ClickOnA provides an extremely e cient -in terms of editing and evaluation time - environment for ontology design.
This paper is structured as follows: Chapter 3 outlines how DL-LiteA speci c aspects are applied in ClickOnA. The next chapter explains its features, like cooperative design, on-the- y inferring and advantages in ontology visualization. Implementation details of the editor including the applied frameworks are introduced in section 5. In the following chapter an evaluation is described which proves the high e ciency of ClickOnA compared to conventional ontology editors.
work Programme \ICT for Energy Systems" 2011-2014, under the grant agreement Number 287534.
Comparative studies for ontology design tools are provided in Khondoker [
Though the methodology issue is not re ected in this paper, it is
important to understand that application of these tools brings maximal e ect in the
context of a particular methodology, where they are supplemented by process
descriptions, templates and other tools. The methodical approach called Data
Integration Driven Ontology Design, in whose context the ClickOnA editor has
been developed and its application is described in [
As stated above the ClickOnA editor has been developed to design ontologies
on the basis of the DL-LiteA formalism, which belongs to the DL-Lite family
of description logics. Another two logics of this family, DL-LiteF and DL-LiteR
were described by Calvanese et al. in [
DL-LiteA support by ClickOnA
In [
Regarding OWL 2 the two most important aspects of DL-LiteA are on the one hand a restriction that allows range/domain speci cations for functional properties only by connecting concepts with data types. On the other hand, for non-functional properties connecting two concepts, domain and range have to be modeled through axioms using \owl:Restriction". For example the following axioms use subsumption (v), existence quanti cation (9) and inversion ( ) to express the fact that the concept Land Vehicle relates to the class Wheel via the hasWheel property.
(1) (2)
In RDF/XML these statements are speci ed as follows: <owl:Restriction> <rdfs:subClassOf rdf:resource=".../clickOnA.owl#Wheel"/> <owl:onProperty> <rdf:Description>
<owl:inverseOf rdf:resource=".../clickOnA.owl#hasWheel"/> </rdf:Description> </owl:onProperty> <owl:someValuesFrom rdf:resource="...org/2002/07/owl#Thing"/> </owl:Restriction> and <owl:Restriction> <rdfs:subClassOf rdf:resource=".../clickOnA.owl#Land_Vehicle"/> <owl:onProperty rdf:resource=".../clickOnA.owl#hasWheel"/> <owl:someValuesFrom rdf:resource="....org/2002/07/owl#Thing"/> </owl:Restriction>
In fact if axioms are speci ed in this style, domains and ranges of the corresponding properties can be inferred by a reasoner software. ClickOnA was implemented to infer ranges and domains from axioms coded this way, converts them into a graphical presentation and thus hides the complex code from the user (Figure 1). Apart from interpretation and graphical representation of axiomatically speci ed domains and ranges, ClickOnA allows generation and modi cation of existing axioms. Here as well, the user designs without the complex axiomatic nature of the speci cation and invokes corresponding operations by selection of a context menu option, as shown in gure 2. Code that strictly aligns with the regulations of DL-LiteA is generated automatically. The editor's on-the- y reasoning functionality enables that this code is interpreted as domain/range speci cations.
De ning DL-LiteA conform code as shown above is probably the biggest problem when a conventional ontology editor like Protege3 is used. To specify one domain/range pair users have to switch between three di erent subwindows: \class description", \object property hierarchy" and \general class
axioms". When a high number of suchlike relations have to be speci ed, errors are almost inevitable. 4
Furthermore ClickOnA implements a set of features aiming at high usability and e ectiveness of the authoring/design process. These features basically are { Partitioning of the complex editing task into a group of smaller and simpler tasks { Advanced ontology visualization, and support of user navigation along the ontology graph 4.1
Partitioning of the editing task ClickOnA considers the editing process as a combination of three sub-processes: i) editing of a subsumption taxonomy; ii) editing of a non-subsumption properties graph and iii) annotation of concepts. Correspondingly, three di erent perspectives, presented in the screen simultaneously are o ered for users: the taxonomy perspective ( gure 3, left), the properties graph ( gure 3, right), and the meta data perspective ( gure 3, upper part).
In the middle column between the taxonomy perspective and the properties graph there is a control panel that can be moved horizontally, enlarging one of the perspectives to the left or to the right side. The advantages of this visual structure become evident when ClickOnA is used on a system with two monitors. In this case one monitor can be dedicated to each of these two perspectives.
A new concept can be created either in the taxonomy perspective or in the properties graph. In the former case this is done using the context menu, when the concept that is supposed to be subsumed by a new one is clicked with the right mouse button. In the property graph, new concepts can be created through a context menu as well. In this case however, instead of a new concept, a new property relation speci ed by a set of axioms, has to be created as described in section 3. The label for this property is generated automatically. New concepts added in the property graph are initially subsuming the concept called Suspense, which is automatically generated by ClickOnA as root of a temporary sub-taxonomy. Later, concepts subsuming Suspense are relocated in the taxonomy to a more suitable place.
Moreover, the meta data perspective (see gure 4) o ers several elds for a concept's metadata. The elds \Comment", \Author", \Reference" and \Label" can be modi ed. Other elds contain values generated by the system. The initial value of the \Author" eld is generated automatically depending on user settings.
Ontology visualization and navigation support To achieve high usability and optimized ontology authoring it is important that users are able to navigate through both inferred graphs (taxonomy and the properties graph) in a most natural and transparent way. In this context ClickOnA o ers the following settings that are shown on the control panel displayed in the middle of gure 3.
{ All nodes of the ontology graph are colored with regard to their meaning (the legend on the control panel explains the di erent colors) { Mouse click navigation: After a concept is clicked on with the left mouse button it is going to be centered on the screen and relations originating from the corresponding graph node and (optionally) form its sub nodes become visible (see node Compactness Ratio in gure 5). Double clicking on a node results in a new tree presentation where the selected concept becomes the root of a taxonomy. (see node CS Geometry in gure 5) { Enlarging or collapsing parts of the ontology graph by a zoom function { Controlling the tree depth (one, two or three levels can be displayed simultaneously) { Changing graph direction from e. g. left to right or bottom down { Forward and backward navigation within a selection history
Additional navigation options can be found in the meta data perspective. There, all super- and sub-concepts of a selected node are listed. By clicking on one of them the editor navigates to the corresponding root node. Besides, the ontology editor o ers a powerful option for searching concepts using autocomplete functionalities. If only a part of a concept name is entered by the user the system o ers him a list of concept names containing the entered part (see gure 6).
In the current version of the ClickOnA editor is able to import and export ontologies as .owl les using the RDF/XML format. Implementation of ClickOnA has been carried out on the basis of a Java Swing graphical user interface and using basically three frameworks: Prefuse4, Apache Jena5 and JDOM 6. The Prefuse framework expands the Java 2D graphic library and can easily be integrated into any Java Swing software. The functionalities of Prefuse are used in ClickOnA to display graphs of concepts and relations. The Apache Jena framework o ers a lot of tools and libraries for reading, processing and writing RDF data in XML; for dealing with OWL ontologies and for rulebased reasoning. JDOM is an open source Java-based document object model for XML. In the ClickOnA JDOM is used to manage (access and modify) les written in the Web Ontology Language (OWL), whereby the OWL is a derivate of the XML, de ned by means of a set of name spaces. 6
Evaluation of DL-LiteA capabilities of ontology editors
The available evaluations of ontology editors (for example [
The advantages of ClickOnA for editing DL-LiteA conform ontologies can be illustrated by a comparison of design processes carried out with ClickOnA and with di erent conventional ontology editors. The editors used for this evaluation are: ClickOnA, Protege 4.2, NeoOn Toolkit 2.5.27, OntoStudio 3.2.08, Hozo 5.2.369, OWLGrEd 1.6.010 and TopBraid Composer 4.2.111.
This evaluation was carried out by three ontology experts. Taking into account that the ClickOnA editor is designed especially for domain experts an evaluation with participants of this group would most probably show a much higher di erence between the process time results, in favor of ClickOnA.
First the process of creating a new, tiny ontology containing 8 concepts has been evaluated. In the second step three properties with axiomatically de ned ranges and domains were speci ed. The third step aimed at presentation of information inferred from axioms created in the previous step. The results of this evaluation regarding the duration of these three processes are displayed in the following table:
Besides ClickOnA only Protege was able to create the axioms required by the DL-LiteA formalism. The creation of new ontology concepts using ClickOnA results in a slightly longer time span because of its current implementation whereby renaming of newly created concepts requires an additional mouse click compared to the other editors. In Protege users have to enter two axioms (see section 2) for each relation manually, whereas ClickOnA generates these fully automatically. On-the- y inferring supported by ClickOnA enables immediate graphic presentation of relations created in the second process (line 3 in the table above). When
10 http://owlgred.lumii.lv/index.html 11 http://www.topquadrant.com/products/TB Composer.html using Protege, at rst a reasoner, for example Fact++, has to be executed before any inferred information can be shown to the user. Unlike ClickOnA Protege isn't able to present this information graphically but only as additional textual description of a selected property ( gure 7).
However the advantages of ClickOnA become evident when the size of the ontology to be edited grows. Comfortable navigation instruments o er perfect orientation, save time and reduce the number of errors signi cantly. 7
In this paper we have presented an innovative ontology editor. ClickOnA is dedicated to the design of ontologies basing on the DL-LiteA formalism. Due to its restrictive architecture application of DL-LiteA may lead to an increasing complexity of the ontology code. In this context real time evaluation of user input becomes essential. Due to on-the- y inferring, powerful graphical presentation and control options ClickOnA is able to give very expressive and illustrative feedback to the user, immediately after his input. Furthermore graphical widgets enable automated rapid and error-free generating of complex axioms, and o er a high level of usability. In the nearest future ClickOnA will be enhanced by mechanisms for versioning, user and change management.