Graphic Relations and Syntactic Structures

We use the term graphic relation5 to refer to the things that we can vary in a visualisa- tion including attributes such as colour and relative relations between graphic objects such as containment . Bertin used the term visual variable for this [Ber83]. Mackinlay puts a basis set of primitive graphical languages that consists of positional language categories such as single position and apposed position and the retinal variables list, but also of the cate- gories map, connection (trees, networks) and misc (pie charts, Venn diagrams). However, when we consider that both attributes and complex, n-ary spatial relations can be used to visualise data, terms like variable or language seem misleading to us. Visual variable could also be confused with visual attribute. Also the terms perceptual tasks [CM84, Mac86a] and visualisation primitives [AA06] we consider problematic, since they may be mixed up with the users tasks or elementary graphic objects. We, therefore, use graphic relation instead.

Graphic Attribute

http://purl.org/viso/graphic/GraphicAttribute

Graphic attributes represent inherent attributes of graphic objects such as colour , size and shape. These attributes can be distinguished from relations between graphic objects (viso-graphic:GraphicObjectToObjectRelations). In addition to the graphic attributes them- selves (e. g., viso-graphic:color_named, possible named values such as viso-graphic:Red and viso-graphic:Green or viso-graphic:Bright and viso-graphic:Dark for lightness are stored in the graphic module.

Examples of graphic attributes that are currently formalised in VISO are given in Fig. 5.6. Superclasses: viso-graphic:GraphicRelation

Subclasses: viso-graphic:GraphicAttributeContinuous, viso-graphic:GraphicAttributeDiscrete Graphic attributes are properties of graphic objects and represent their inherent characteris- tics such as colour , size and shape. Following Engelhardt, we distinguish graphic attributes, which characterise a graphic attribute in a graphic space, and relations between graphic ob- jects, which will be described below. Various variants of the term graphic attribute have been used in literature for similar, yet often not identical, concepts: Bertin introduced the term retinal variables for properties to which the retina is sensitive independent of the movement of the eye: size, saturation, texture, colour , orientation, shape. Other categories of Bertin are positional and temporal variables. Later authors reused this classification [Mac86a, Maz09]. Similarly Engelhardt classifies visual attributes into two groups: spatial attributes and area-fill attributes [vE02]. He describes the difference as follows: »If we would regard every point of a graphic object as being anchored to its location in graphic space, then varying a spatial attribute of the object would alter this anchoring, while varying an area-fill attribute of the object would not alter this anchoring.« Andrienko and Andrienko again use the term retinal variables that they see as »internal, individual properties of marks« and distinguish them from dimensions

4 _{Lines – from a perception point of view – cannot actually be 1-dimensional, but must be thin areas too, in}

order to be visible.

5 _{Up to this point, we sometimes used the less technical visual means instead of graphic relation. The}

CHAPTER 5. A VISUALISATION ONTOLOGY – VISO

(a) Colour (named). (b) Shape (named). (c) Texture (named).

(d) Hue (HSL). (e) Saturation (HSL). (f ) Lightness (HSL).

Figure 5.6: Examples of graphic attributes defined in the VISO/graphic module. a–c are examples of discrete »named« attributes, i. e., they will point to fixed values that are stored as resources in the ontology as NamedGraphicAttributeValue, e. g., viso-graphic:Red. Subfigures d–f show examples of continuous attributes, which point to a value that can be stored as an RDF literal.

5.5. VISO/GRAPHIC MODULE – GRAPHIC VOCABULARY

which they describe as »containers for marks«. Not only spatial position, but also time plays an extra role among the visual attributes, since both are physical dimensions or spatial attributes. Although time is often less discussed, already Bertin introduced temporal relationships and gave animation as an example. Besides using time as another physical dimension, it can also be used

in attributes such as flicker frequency .

Finally, it is important to note that some of the graphic attributes have multiple dimensions, or are bundled (as Wilkinson [Wil05] calls it). Position belongs to this category of attributes (three spatial dimensions) but also colour may be seen as having multiple dimensions that span a colour space (e. g., hue, saturation, lightness or red, green, blue). In VISO/graphic, we model each of these dimensions as a separate property, i. e., we have three properties viso-graphic:color_hsl_hue, viso-graphic:color_hsl_saturation and viso-graphic:color_hsl_lightness.

For graphic attributes that can take continuous values from a specific value range, the (maximum) value range can be stated using viso-graphic:max_value_range, which is neces-

sary to enable default value mappings (Sect. 7.7). For example, VISO/graphic defines that viso-graphic:color_hsl_lightness can range between »0« and »100«.

Graphic Object-to-Object Relation

http://purl.org/viso/graphic/GraphicObjectToObjectRelation

The class of all properties that are visual relations between complex objects. Visual relations build visual (syntactic) object-to-object structures.

Examples of (elementary) graphic object-to-object relations that are currently formalised in VISO are given in Fig. 5.7 (a–i).

Superclasses: viso-graphic:GraphicRelation

Subclasses: viso-graphic:BinaryGraphicO2ORelation, viso-graphic:N-AryGraphicO2ORelation viso-graphic:GraphicObjectToObjectRelation is a second subclass of viso-graphic:GraphicRelation besides viso-graphic:GraphicAttribute, cf. Fig. 5.4). To explain the origin of this concept, we now have a close look at the aforementioned »set of graphic relations« and the resulting graphical syntactic structures, as Engelhardt calls them [vE02]. Since he distinguishes spatial and area-fill attributes, he introduces spatial syntactic structures and attribute-based syntactic structures. He further distinguishes structures of relations in-between graphical objects and between graphical objects and graphical space. Structures involving object-to-object relations are, for example, spatial clustering, linking , containment and superimposition. Card calls them topological structures [CMS09]. An example for a structure involving object-to-space relations

is a coordinate system that spans a metric space.

Andrienko and Andrienko’s dimensions are similar to Engelhardt’s spatial syntactic structures. They are concerned with everything that provides a position. This includes physical dimensions, but also various arrangements of the display space. Examples of arrangements are node-link- structures and discontinuous tables that resemble Engelhardt’s spatial object-to-object structures. The treatment of graphic relations and spatial structures is the field where we encountered the most diverging approaches. Engelhardt’s description of object-to-object and object-to-space structures as well as the consideration of both attribute- and spatial-relationships covers all our use cases and uses coherent terms. For this reason, we picked his terms for reuse in VISO to a large extent, including the notions of object-to-object relations, which we model as viso-graphic:GraphicObjectToObjectRelation.

In Fig. 5.7, besides the elementary graphic object-to-object relations (a–i), the subfigures j–l show further relations between graphic objects, which are more complex and need to be composed from elementary relations. For example, labelling needs to be »realised« using a connector or drawing the label (relatively) close to the labelled object as shown in the figure. Also building and co-highlighting include elementary relations or attributes such as line-up,

CHAPTER 5. A VISUALISATION ONTOLOGY – VISO

position and colour. The formalisation of the exact relation between elementary and »composite« graphic relations is subject to further research. For example, currently VISO does not allow for specifying how labelling should be realised. This may cause interactions with graphic relations chosen in other mappings.