Information Technology Reference
In-Depth Information
Many areas of research have highly developed conventions and standards for
diagrams. Browsing relevant papers in the same area as your work should you give
a good idea of what elements a diagram should incorporate and of how it should be
presented.
Broadly speaking, diagrams are used to show either a structure, a process, a rela-
tionship, or a state. Although these are high-level distinctions, they are valuable
because a common mistake in design of diagrams is to attempt to combine these pur-
poses inappropriately. For example, a schematic showing data flow in an architecture
is likely to be unclear if control flow is also illustrated.
Some forms of diagram and illustration are automatically generated by tools
from data. In particular, mechanisms for data visualization can be used to build rich
images. Here, however, I am primarily concerned with the line drawings that form a
key part of many papers. While automatic tools for generating diagrams can be used
to produce a wide variety of representations from the same underlying data, some of
which will be dramatically more effective than others, the richness and diversity of
these tools—not to mention the rate at which they are developing—means that such
diagrams are beyond the scope of this topic.
To design a diagram that is to be created with a manual tool, the first step is often
to do initial sketches by hand, on paper. This early stage is the appropriate time to
balance the diagram, by checking that it is well-proportioned, makes good use of
the space, is laid out well and doesn't have the elements bunched to one side, and is
arranged so that the relative sizes of the elements look reasonable. However, never
submit a paper with a hand-drawn diagram unless it has been prepared by a profes-
sional; almost any diagram can be drawn well with the tools available for a typical
computer.
A diagram should not be too dark; keep it as sparse as possible. This is best
achieved by eliminating all clutter. A diagram does not have to be faithful to every
detail of the concept being illustrated; fine details can always be clarified in the
supporting text and even the best diagram requires some explanation. Use meaningful
labels, which should if possible be displayed horizontally, and make the point size
and font of the labels similar to that of the other text. As for text in general, there
should be no more than two or three fonts and font sizes.
Lines should not be too heavy, but at most a little thicker than the lines used to
draw the text font. Shades of grey can be used to distinguish between solids but are
not as effective for distinguishing between lines, and don't use shades that are too
light or too similar to each other. Pictorial elements should be used consistently, so
that, for example, arrows and lines of the same kind have the same meaning. Use
shading rather than cross-hatching. Colour can be highly effective, especially if it is
used sparingly; but, as for graphs, do not use colour if the paper will ultimately be
printed in black-and-white, or if the sole reason is to make the paper more attractive
to look at. If arrows are used to show arcs as well as to point at features, distinguish
them by, say, using dashed lines in one case and solid lines in another. Lines should
not touch each other unless separating them would create an unnatural break. Thus,
for example, there should usually be a gap between an arrowhead and the element
the arrow is pointing at.
