A second group of requirements concerns the technical aspects of the ontology. Since we aim at using the VISO vocabulary to describe the »target side« of our explicit, declarative mapping definitions (R-9), we also need to make visual means explicit (VR-6). Moreover, directly using the defined visual means for a (guided) configuration of graphic representations (R-14, R-13 and R-15) requires us to store both the graphic vocabulary and the visualisation knowledge in a formal, machine-processable and understandable way (VR-7). The criterion of platform variability (R-10) enforces the description of visualisation terms and knowledge to be platform- independent (VR-8) as well. To allow for mapping the specifics of ontologies (R-16), it has to be simple to reference ontological terms within these rules (VR-9).
V0-1 The models should be easily accessible, preferably by the same technologies as the data to be visualised.
Furthermore, the ontology should be easily processable by the same technologies as the data being processed and should be easily shared and integrated with existing vocabularies from the Semantic Web community (V0-1; optional requirement). This suggests building the ontology in the widespread and standardised RDF-based Web Ontology Language (OWL) to avoid technological gaps.
5.3
Existing Approaches to Modelling in the Field of Visu-
alisation
Numerous models have already been developed in the field of visualisation with different goals, e. g., to classify and clarify concepts of the visualisation domain, to describe the visualisation process and even to model knowledge for visualisation systems. Already in 1986, Mackinlay started to formalise visualisation knowledge as LISP-rules [Mac86a]. Furthermore, there are multiple taxonomies on visualisation stressing different aspects of it, such as interaction, tasks or the characteristics of underlying data. In preparation of the VISO ontology, we compared a broad corpus of articles from the field of graphics and visualisation, especially those already suggesting terminologies, taxonomies and ontologies. The objective was to comprehend which fields of visualisation are covered and whether an ontology already exists that could be reused for our purposes. Further, we needed to understand what are the drawbacks of existing models and to identify important work, which could serve as a basis for the VISO ontology.
For classification models, various level of formalisation are possible. Duke et al. [DBDH05] distinguish the following three levels of formalisation for classifications: Terminologies intro- duce concepts in a less structured, informal way. Taxonomies define concepts in a hierarchically structured but mostly informal way. Ontologies are the most formal approach where concepts and their relations are based on a shared meaning. Since we require a model that can be equally processed and understood by humans and machines, a high level of formalisation is a critical factor. However, although a few initial ontologies existed in the visualisation domain, they did not cover our requirements (Sect. 5.3.2). Therefore, we extensively compare terminologies and taxonomies as well, of which many are described in the literature of the last decades (Sect. 5.3.1). Finally, we summarise other visualisation models (Sect. 5.3.3) whose main purpose
is not classification.
5.3.1
Terminologies and Taxonomies
In the domain of visualisation and related areas, numerous terminologies and taxonomies have been developed with different goals, e. g., to allow for systematic reviews of existing techniques
CHAPTER 5. A VISUALISATION ONTOLOGY – VISO
and ideas. It is not possible to discuss them all in depth in this work, hence, Fig. 5.3 illustrates the overall results of surveying 53 articles. An extensive comparison can be found in the Appendix B.1. We distinguish the field of data and the domain of the data. Further, we refer to graphic representation as the result of the visualisation process, which is synthesised using a graphic vocabulary. We subsume the topics task and interaction as activity (cf. Sect. 5.8). The fields user and system are about modelling the user and system context and how this benefits the visualisation process.
From our survey, we discovered three findings. In Fig. 5.3-a, one can observe that most classifications concentrate on data, graphical representation and activity. The other fields got less attention and, thus, seem to be good future directions of research. Further, only few works try to unify a broader spectrum of concepts of the interdisciplinary domain of visualisation, because in most cases only one or two, sometimes three areas are tackled by a single work (cf. Fig. 5.3-b). Finally, about 90% of the reviewed literature deals with terminologies or taxonomies and only 10% deals with ontologies, which we have a closer look at in the next section.
0% 10% 20% 30% 40% 50% 60% 70% a) b) 0 5 10 15 20 1 2 3 4 5 6 7 Da Do Vo Re Ac Us Sy
Figure 5.3: Statistical overview of reviewed literature (a) by fields they focus in general [Data (Da), Domain (Do), Graphical Vocabulary (Vo), Graphic Representation (Re), Task and Interaction (Ta), User (Us) and System (Sy )] and (b) by how many areas are covered within a single work.
5.3.2
Existing Visualisation Ontologies
In the following, we give an overview of existing visualisation ontologies and relate them to our objectives. Based on a workshop held at the National e-Science Centre of Scotland in 2004 [BDD+04], Duke et al. advise to merge the existing visualisation knowledge fragments by means of an ontology [DBD04]. They provide a vocabulary by which users and system can communicate. It comprises only a small set of concepts and relations like data, visual representation and task. Unfortunately, an implementation or a more detailed version of their ideas are missing. Potter and Wright [PW07] took up the idea for the description of visualisation resources with focus on hard- and software requirements. Their ontology is not accessible and also misses a comprehensive overview. Rhodes et al. [RKR06] worked on an application to categorize and store information about software visualisation systems. Although they state to incorporate the concepts of Duke et al. [DBD04], the paper lacks a clear description of this fusion. Further, the developed ontology schema is tailored to software visualisation, thus, the work does not directly contribute to a domain-independent visualisation ontologies. The Visual Representation Ontology that Gilson et al. [GSGC08] propose as part of their tool for automatic visualisation, comprises properties of the entire graphic representation as well as attributes of single graphic objects. While this work is promising in terms of formalisation, its focus is narrow: Neither interaction or tasks nor the user are considered in this ontology. Finally, Shu et al. [SAR08] created an ontology for visualisation, intended for the semantic description of visualisation services. Based on the initial visualisation ontology [DBD04], as well as on taxonomies proposed in [BCE+92] and [TM04b], their visualisation ontology mainly comprises classes for modelling data and visualisation techniques. However, they did not consider concepts