=Paper=
{{Paper
|id=None
|storemode=property
|title=Multi-Attribute Glyphs on Venn Diagrams to Represent Quantities and Aid Visual Decoding
|pdfUrl=https://ceur-ws.org/Vol-854/paper9.pdf
|volume=Vol-854
|dblpUrl=https://dblp.org/rec/conf/diagrams/Brath12
}}
==Multi-Attribute Glyphs on Venn Diagrams to Represent Quantities and Aid Visual Decoding==
https://ceur-ws.org/Vol-854/paper9.pdf
Multi-Attribute Glyphs on Venn and Euler Diagrams to
Represent Data and Aid Visual Decoding
Richard Brath
Oculus Info Inc., Toronto, ON, Canada
richard.brath@oculusinfo.com
Abstract. Representing quantities on Venn and Euler diagrams can be achieved
through the use of multi-attribute glyphs. These glyphs can also act as an aid to
assist in the visual decoding of the membership of segments within the
diagrams and convey other data attributes as well.
1 Overview
Instead of area-proportional Venn and Euler diagrams to indicate quantities, this
approach uses separate overlaid glyphs to decouple the representation of data from
logical combinations. It also uses glyph attributes to assist in visual decoding of
membership of regions. This approach can scale to higher-order logical diagrams and
potentially offer more accurate visual estimation than area-proportional techniques.
The depiction of data on set diagrams is useful in various applications (e.g.
Boolean queries, genetic informatics). Area-proportional set diagrams have become
popular in research (e.g. [1-6]) and software (e.g. see eulerdiagrams.org). However,
the area-proportional approach has shortcomings, such as:
1. Visual comparison of irregular areas is difficult. Information visualization
researchers indicate difficulty with visual comparison of areas and/or a preference
for using length instead of area for faster visual comparison (e.g. [7,8,9,15]). In our
casual test, only 8% (2 out of 25 people) correctly identified the region of different
area on a 2 way Venn as opposed to 80% correctly identifying the circle of
different area out of three circles, each test having one item of 20% different area.
2. Area accuracy vs. aesthetic shapes. Researchers prefer circles and other aesthetic
shapes[6], but the areas (particularly circles) may have a degree of error, typically
increasing with higher order sets. e.g.[1]. Wilkinson [6] says “Higher-order Venn
diagrams can be drawn on the plane with nonconvex polygons, but they are
difficult to compute for more than a few sets and are difficult to decode visually.”
3. Negative values: Areas cannot represent negative values unless coupled with
another visual attribute, such as hue (e.g. red/green) or shape (e.g. arrows).
Also, discussions with prospective users revealed concern for visual decoding of set
membership for a region in complex diagrams, such as higher-order Venn diagrams.
3rd International Workshop on Euler Diagrams, July 2, 2012, Canterbury, UK.
Copyright © 2012 for the individual papers by the papers' authors. Copying permitted for private and
academic purposes. This volume is published and copyrighted by its editors.
123
Instead, a glyph-based approach is considered. The use of glyphs within set dia-
grams is not new. Glyphs have been used to represent items in a dataset (first 3 in fig.
1). Spoerri’s approach [10] reduces each region of a Venn diagram to a glyph, each
glyph indicating the particular Boolean combination by its relative position and shape.
Fig. 1. Top Left: TwitterVenn [11] uses simple glyphs per tweet matching each of three search
terms. Top Right: Edwards’ Carroll diagram [2] uses both proportional areas (to indicate
expectations) with dots (to indicate observations) to draw attention to regions with unexpected
results. Bottom Left: [12] depicts a uniform density of glyphs with membership indicated by
shading or outlined boxes. Bottom Right: InfoCrystal [10] reduces each region of a Venn
diagram to a glyph, where shape and position indicate the Boolean relationship, and the number
indicates the count of items within each region.
The contribution of this paper explores, in section 2, the use of glyphs
(pictographic and scaled glyphs) to indicate quantities and the use of additional visual
attributes to indicate set membership or other data. Results are discussed in section 3.
2 Glyph-based Approach
Our approach is focused on the use of glyphs to decouple the depiction of logical
relationships (e.g. Venn and Euler diagrams) separate from the depiction of quantities.
By decoupling the quantity from the set diagram, visual attributes more amenable to
fast estimation (e.g. size, origentation and color) can be used [7-9]. This approach
enables the use simple aesthetically pleasing diagrams of set representations to show
the logical relationships between the sets; while using separate glyph(s) within each
region to indicate a) quantity of items within a given region, b) indicate set member-
ship to aid visual decoding and c) potential additional attributes.
124
2.1 Sketches and Real-Data Mockups
To quickly iterate through conceptual ideas, loose sketches were followed by mock-
ups using simple sets of real-data. Loose sketches can reveal limitations of promising
ideas when implemented with real-data, (e.g. occlusion, imperceptible differences,
large dynamic ranges, etc). For rapid mock-ups, we divided the Titanic passenger list
into 4 sets for a Venn diagram and 3 sets for an Euler diagram which resulted in use-
ful properties such as empty segments, small segments and large segments.
Fig. 2. Left: Venn diagram of Titanic passengers by 4 attributes showing passenger counts.
Right: Euler diagram of Titanic passengers and crew.
2.2 Unit Markers and Pictographs
Markers of a fixed size can be repeated to represent quantities ranging from simple
dots to pictographs e.g. Isotype [13]. Additional data can be represented on each
marker, e.g. using color or sub-shapes. For example, “Social Stratification in the
United States” [14] uses pictographic markers with human figures indicating five
variables, through 1) background color (occupation), 2) shape (gender), 3) pairing
(marital status), 4) extra outline (dependents), and 5) figure color (race). We have also
used this approach successfully, e.g. [15].
Fig. 3. Pictographic glyphs indicating multiple data variables, from [13,14].
However, a pictograph approach has some challenges:
• Some regions of the set diagram are too small to fit the pictographs. The addition
of the leader line could increase the effort to visually decode the relationships.
• The irregular shape of some regions of the set diagram requires an irregular place-
ment. Pictographs organized linearly can be visually estimated by length, which is
preferred to visual estimation of area (e.g. [8]).
125
Fig. 4. Pictographs on set diagrams. Leader lines are used when pictographs do not fit regions.
2.3 Scalable Glyphs
Scalable glyphs are a single glyph for each region, sized by the quantity of items as-
sociated with that region. Simple glyphs, such as bars varying in length or circles
varying in radius, can effectively convey quantities [8,9,16]:
Fig. 5. Scaled glyphs indicating quantities on both Venn and Euler diagrams.
2.4 Indication of Set Membership
With higher order set diagrams, it can be more difficult to perceptually decode the
membership for a given component of interest [6]. There are many possible approach-
es to indicate set membership using either the set diagram or the glyphs.
Background Color: Color can be used, but is problematic. It is challenging to de-
code the color in intersections as color is not understood as separable [16].
Background Texture: Textures have been used to aid in identification of set
membership e.g. [17, 18]. Distinguishing regions by a heterogeneous channel-based
approach [19] could be more effective. However, in small regions, textures may not
be clearly distinguishable or the glyphs may occlude textures.
126
Fig. 6. Background textures can assist in decoding membership.
Glyph with Colors: The same coloring used in the set diagram can be reused in
glyphs to indicate membership. Rather than blend colors, however, the colors can be
kept separate within the glyph. The layout of the color could be organized as stripes,
or radially resembling a bullseye or pie.
Fig. 7. Colored glyphs use same colors as the sets to aid in identification of set membership.
Glyph with Oriented Whisker: In some set diagrams (e.g. Venn) the placement
of the label is typically around the perimeter which can be leveraged by the glyph,
specifically by modifying the shape with an added whisker [20] oriented along the
same vector from the center of the diagram to the label. To visually decode the mem-
bership of any bubble, the viewer can read the orientation of the whisker, similar to
decoding the hands of a clock or the spokes of a wind rose.
Fig. 8. Glyphs with added whiskers oriented based on set memberships.
The glyph-based approach also allows for additional visual attributes to convey addi-
tional data values. For example, the whisker-based glyphs can use:
127
• Traditional visualization attributes, such as brightness, hue, texture
• Shape-based attributes, such as closure, curvature or edge type
• Each whisker-shape can be independently modified to indicate a data attribute with
respect to the set membership, for example whisker length or width
• The internal area of the glyph can be used, in larger glyphs, for example, as a pie
chart or with a pictograph.
Fig. 9. Example whisker glyphs with additional visual attributes, a) brightness, b) closure,
c) curvature, d) whisker width and length, e) internal pie, f) internal pictograph. Negative val-
ues could be connotatively conveyed by fill (e.g. red/green hue) or pictograph (up/down arrow).
The whisker-based approach may work well with Venn diagrams, but has may have
issues with Euler diagrams and issues where whiskers are potentially occluded. The
image below shows a 5-way Venn diagram that has been modified from ellipses to
increase the size of the smaller regions to make the technique more workable.
Fig. 10. Whisker glyphs on a 5 way Venn diagram showing data from a survey of 5000 people
purchasing fuel, in sets by age, payment type, fuel grade, residence and income. Size of glyph
indicates number of respondents in a given component, whiskers indicate set membership, and
angular shading indicates the ratio of female to male purchasers of a component.
128
3 Discussion and Next Steps
Our contribution shows that glyphs can be used to separate the representation of data
such as quantities from the representation of sets. Glyphs can:
• Indicate data layered over set diagrams, either as scalable glyphs or as pictographs,
and simple size is preferred for fast visual estimation as opposed to irregularly
shaped areas [16].
• Also indicate additional data attributes, such as set membership or other data at-
tributes, using visual attributes such as color or orientation of sub-shapes.
While current research in visual comparison indicates this approach may work, evalu-
ation is required to validate. The examples provided indicate various limitations:
• Glyph size needs to be carefully managed. Glyphs too big can result in occlusion
or require an offset and leader lines. Glyphs too small can be difficult to add addi-
tional visual attributes, e.g. for visual decoding.
• Glyph colors can effectively represent set membership except color can be diffi-
cult to discern when used internally on small glyphs.
• Glyph whiskers can effectively represent set membership when set labels are or-
ganized around the perimeter, such as in Venn diagrams; but is problematic when
sets are distributed throughout the plane.
Implementation on a wide variety of data sets and testing with users requires further
effort. Other work could include 3D and interaction techniques, for example, on inter-
action, extend whiskers to set labels to aid interpretation of whiskers.
References
1. Chow, S., Ruskey, F.: Drawing Area-Proportional Venn and Euler Diagrams. Lecture
Notes in Computer Science 2912:466–77 (2004)
2. Edwards, A.W.F., Edwards, J.H.: Metrical Venn diagrams, Annals of Human Genetics 56:
71-75 (1992)
3. Micallef, L., Rodgers, P.: eulerAPE: Drawing Area-Proportional Euler and Venn Diagrams
using Ellipses. EMEA Google Scholars Retreat 2011 www.cs.kent.ac.uk/pubs/2011/3119
(2011)
4. Chow, S., Rodgers, P.: Constructing Area-Proportional Venn and Euler Diagrams with
Three Circles. Presented at Euler Diagrams Workshop 2005, Paris (2005)
5. Pirooznia, M., Nagarajan, V., Deng, Y.: GeneVenn – a web application for comparing
gene lists using Venn diagrams. Bioinformation, 1:420–422 (2007)
6. Wilkinson, L.: Venn and Euler Data Diagrams. Science (2), Citeseer (2010)
7. Healey, C., Booth, K., Enns, J.: High-speed visual estimation using preattentive pro-
cessing. ACM TOCHI (1996)
8. Jock Mackinlay. Automating the Design of Graphical Presentations, ACM Transactions on
Graphics, 5(2) (1986)
129
9. MacEachren, A.: How Maps Work: Representation, Visualization and Design (2004)
10. Spoerri, A.: InfoCrystal: A Visual Tool For Information Retrieval And Management. In:
CIKM '93 Proceedings of the second international conference on Information and
knowledge management (1993)
11. Clark, J.: TwitterVenn. www.neoformix.com/Projects/TwitterVenn/view.php
12. Brase, G.L.: Pictorial representations in statistical reasoning. Applied Cognitive Psycholo-
gy, 23(3), 369-381 (2009)
13. Neurath, O.: International picture language. London: Kegan Paul (1936)
14. Rose, S.: Social Stratification in the United States: The American Profile Poster. The New
Press (2007)
15. Jonker, D., Wright, W., Schroh, D., Proulx, P., Cort, B.: Information Triage with TRIST.
2005 Intelligence Analysis Conference (2005)
16. Ware, C.: Visual Thinking for Design. Ch. Structuring Two-Dimensional Space,
pp. 43-65, Morgan Kaufmann (2008)
17. Byelas, H., Telea, A.: Texture-based Visualization of Metrics on Software Architectures.
Software Visualization (2008)
18. Simonetto, P., Auber, D., Archambault, D.: Fully Automatic Visualisation of Overlapping
Sets. Eurographcs/IEEE-VGTC Symposium on Visualization (2009)
19. Ware, C.: Information Visualization: Perception for Design. Ch. Glyphs And Multivariate
Discrete Data, pp. 176-184 (2004)
20. Brath, R.: The Many Dimensions of Shape. IV'09 Opening Keynote,
www.oculusinfo.com/expertise.html (2009)