Information Technology Reference
In-Depth Information
Motivation
Many authors take considerable trouble over the structure of their papers but don't
make the structure obvious to the reader. Not only should the parts of a paper be
ordered in a logical way, but this logic needs to be communicated.
The introduction usually gives some indication of the organization of the paper,
by outlining the results and their context, and may include a list of the parts of the
paper, but these measures by themselves are not sufficient. Brief summaries at the
start and end of each section are helpful, as are sentences connecting one section to
the next; for example, a well-written section might conclude with:
Together these results show that the hypothesis holds for linear coefficients.
The difficulties presented by non-linear coefficients are considered in the next
section.
Link text together as a narrative—each section should have a clear story to tell. The
connection between one paragraph and the next should be obvious. This principle is
sometimes expressed as: Tell the reader what you are going to say, then say it, and
then tell the reader that you have said it.
A common error is to include material such as definitions or theorems without
indicating why the material is useful. Usually the problem is lack of explanation;
sometimes it is symptomatic of an ordering problem, such as including material
before the need for it is obvious. Never assume that a series of definitions, theorems,
or algorithms—or even the need for the series—is self-explanatory. Motivate the
reader at eachmajor step in the exposition: explain how a definition (theorem, lemma,
whatever) is to be used, or why it is interesting, or how it fits into the overall plan.
The authors of a paper are almost always better informed than their readers. Even
expert readers won't be familiar with some of the details of a problem, whereas the
author has probably been studying the problem intimately for months or years, and
takes many difficult issues for granted. You should explain everything that is not
common knowledge to the paper's readership; what constitutes common knowledge
depends on the paper's subject and on where it is published. At each part of a paper
you should consider what the reader has learnt so far, whether this knowledge is
sufficient to allow understanding of what follows, and whether each part follows
from what has already been said.
Motivation is essential, but do motivate the right thing. Don't, say, set the scene
by explaining that certain algorithms are too slow for massive databases, and then
test your method on a few thousand records; or argue that “Web search needs to
be semantically aware” and then propose the use of spelling correction to amend
queries. The big-picture topic may well be the inspiration for your overall work, but
that does not mean that it is necessarily the right inspiration for a particular paper.
1
1
And don't let the motivation take over. In a paper on routing of traffic in a local network, the author
was attempting to introduce the topic of congestion measurement, but digressed into a history
of network topology, configuration, scale, and hardware. It took two pages to reach what appeared to
