A Visual Notation for the Integrated Representation of OWL Ontologies
Total Page:16
File Type:pdf, Size:1020Kb
A Visual Notation for the Integrated Representation of OWL Ontologies Stefan Negru1 and Steffen Lohmann2 1Faculty of Computer Science, Alexandru Ioan Cuza University, General Berthelot 16, 700483 Iasi, Romania 2Institute for Visualization and Interactive Systems (VIS), Universit¨atsstraße 38, 70569 Stuttgart, Germany [email protected], [email protected] Keywords: Ontology Visualization, Visual Notation, OWL, Semantic Web, Knowledge Visualization, Ontologies. Abstract: The paper presents a visual notation for the Web Ontology Language (OWL) providing an integrated view on the classes and individuals of ontologies. The classes are displayed as circles, with the size of each circle representing its connectivity in the ontology. The individuals are represented as sections in the circles so that it is immediately clear from the visualization how many individuals the classes contain. The notation can be used to visualize the property relations of either the classes (conceptual layer) or of selected individuals (integrated layer), while certain elements are always shown for a better understanding. It requires only a small number of graphical elements and the resulting visualizations are comparatively compact. Yet, the notation is comprehensive, as it defines graphical representations for all OWL elements that can be reasonably visualized. The applicability of the notation is illustrated by the example of the Friend of a Friend (FOAF) ontology. 1 INTRODUCTION reasoning. Even though it provides great means to de- scribe and integrate information on the Web, its text- The Web of Data has attracted large interest for both based representation is difficult to understand for av- data publishers and consumers in the last years. Sev- erage users. eral organizations have begun publishing their data in While the World Wide Web Consortium (W3C)3 RDF (Resource Description Framework)1 and link- provides a notation for the visualization of RDF ing it with other data, leading to a significant growth graphs4, it does not provide a notation for OWL. in RDF triples available on the Web of Data (Bizer Since any OWL ontology can be represented as RDF et al., 2009). As (Dadzie and Rowe, 2011) point out graph, it can also be visualized using the RDF no- “making sense of such data presents a huge challenge tation. However, as every single RDF statement is to the research community”. User-centered visualiza- separately visualized, this can result in very large vi- tions are considered essential to help people under- sualizations with plenty of nodes and edges. Such standing the data. RDF visualizations of OWL ontologies are not only Ontologies play a key role in this context. An on- hard to read but they also fail to adequately reflect the tology is defined as “a set of knowledgeterms, includ- OWL semantics, as the OWL constructs are split up ing the vocabulary, the semantic interconnections and into multiple RDF triples. some simple rules of inference and logic, for some Therefore, research has addressed the issue of particular topic” (Hendler, 2001). The primary pur- providing visualizations for OWL ontologies. How- pose of an ontology is to classify things in terms of ever, most of the existing approachesrepresent classes semantics for a specific domain. Ontologies provide and individuals in a mutually exclusive manner, with- the backbone for publishing RDF data in a semanti- out providing an integrated view. Moreover, differ- cally meaningful way. Especially the Web Ontology ent graphical representations are used to depict the Language (OWL)2 is widely used to define the con- classes, individuals, properties, and their relation- ceptual schema, the classes, individuals, and the re- ships. An integrated view on ontologies that uses a lationships between them. However, OWL is mainly uniform graphical representation can therefore be re- designed for use by applications in order to support garded as crucial for a better understanding of ontolo- interoperable data exchange, machine processing, and gies and linked data. 1http://www.w3.org/RDF/ 3http://www.w3.org 2http://www.w3.org/2004/OWL/ 4http://www.w3.org/RDF/Validator/ 308 AVisualNotationfortheIntegratedRepresentationofOWLOntologies In order to fill this gap, this paper presents a vi- OWL ontologies. It has been particularly designed for sual notation for OWL ontologies that represents both the sublanguage OWL DL and the semantics of De- classes and individuals in a comprehensive yet com- scription Logic (DL), for instance, by offering sepa- pact manner. A central goal is to provide both expert rate visualizations for the TBox, ABox, and RBox of and non-expert users with an easy-to-understand no- an ontology (Rudolph, 2011). However, the graphical tation optimized for OWL ontologies. The key idea representation quickly becomes complex and large in of the approach is to offer an integrated view on on- size, as every individual is defined in a separate node. tologies, where classes are represented as differently Work has also been done in the direction of sized circles that are shown along with the individuals representing ontologies as UML class diagrams8, they contain. mostly by defining mappings between elements from In the following, we first summarize related work OWL and UML (Cranefield, 2001). The Object in Section 2. In Section 3, we present the developed Management Group (OMG) formally specified such visual notation for OWL ontologies. In Section 4, we mappings in the Ontology Definition Metamodel illustrate the applicability of the notation by the exam- (ODM)9. One implementation in this context is OWL- ple of the Friend of a Friend (FOAF) ontology. The GrEd (Barzdins et al., 2010), a graphical OWL ed- paper ends with a conclusion and an outlook on future itor that uses UML elements for the representation work in Section 5. of ontologies. A disadvantage of these approaches is again their limited visual scalability, especially when it comes to the representation of a large number of individuals. 2 RELATED WORK Other approaches focus on more specific visual- ization tasks. For instance, the RelFinder (Heim et al., Several approaches for the visual representation of 2010) visualizes relationships between individuals in ontologies and linked data have been presented in the ontologies and makes these relationships interactively last years (Guo and Chan, 2011; Dadzie and Rowe, explorable. It nicely depicts a limited set of instance 2011; Katifori et al., 2007; Geroimenko and Chen, data and helps to discover commonalities, but it does 2006). Most of them are based on a two-dimensional not provide a global picture of the ontology. plane, though some propose to use 3D representa- OntoTrix (Bach et al., 2011) presents a hybrid net- tions. Little surprising, most users “seem to dislike work visualization technique that uses both node-link chaotic and too cluttered overviews” but “tend to pre- diagrams and adjacency matrices to represent parts fer visualizations that offer the possibility of an or- of the ontology. Though an interesting combination derly and clear browsing of the presented informa- of visualization techniques, it relies on the users’ ca- tion” (Katifori et al., 2007). Finding a good balance pabilities to correctly interpret the matrix representa- between comprehensiveness and ease-of-use is one of tions and the used color-coding, which can be prob- the key challenges. Therefore, many works show ei- lematic at least for average users. ther only a part of the ontology or end up with very Finally, there are also visualization approaches complex representations. that represent the classes and individuals in ontolo- A number of visualization approaches have been gies as differently sized circles. One such work is implemented as plugins for ontology editors like CropCircles (Wang and Parsia, 2006), which visu- 5 6 Prot´eg´e . Some of them (e.g. OWLViz ) simply visu- alizes the OWL class hierarchy using nested circles. alize the class hierarchies of ontologies, while others The size of the circles is proportional to the position 7 (e.g. OntoGraf ), allow for more comprehensivevisu- of the classes in the hierarchy. OOBIAN Insight10 alizations that include a representation of individuals also makes use of nested circles to visualize ontology- and different property relations. Yet, no clear distinc- based data. Like in GrOWL, the visualization quickly tion between property types or between classes and becomes complex and large in size, making it hard individuals is made in the graphical representation of to read. Another example of visualizing ontologies OntoGraf. The notation rather relies on colors and with nested circles is the DOPE Browser presented in abstract symbols to indicate different node and link (Stuckenschmidt et al., 2004), where individuals are types. grouped into circle-shaped class representations and A more complete OWL visualization is provided linked together to form a cluster map. by GrOWL (Krivovet al., 2007),a graphicaleditor for The review of related work showed that most ex- 5http://protege.stanford.edu 8http://www.uml.org 6http://protegewiki.stanford.edu/wiki/OWLViz 9http://www.omg.org/spec/ODM/ 7http://protegewiki.stanford.edu/wiki/OntoGraf 10http://dbpedia.oobian.com 309 WEBIST2013-9thInternationalConferenceonWebInformationSystemsandTechnologies isting approaches focus on the terminological part of ontologies (i.e. the classes and their relationships), 3 instances + 2 inferred while little emphasis is placed on the individuals be- instances longing to