Information Technology Reference
In-Depth Information
savings if all the data fits in memory. For large tables, on the other hand, the savings
due to reduced disk traffic, increases in the numbers of records per block, and use
of less temporary space may be significant. Thus it seems likely that the savings due
to compression would increase with the size of table to be sorted, suggesting that
results be presented in a graph of data volume against sorting time for fixed block
size. Note too that the question of what to measure identifies an implicit assumption:
that the data was uncompressed to begin with and is returned uncompressed. All of
these decisions and steps help to determine the paper's content.
You may be reporting a particular piece of work, but the way it is reported is
determined by the characteristics of the audience. For example, a paper on machine
learning for computer vision may have entirely different implications for the two
fields, and thus different aspects of the results might be emphasized. Also, an expert
on vision cannot be assumed to have any expertise in machine learning, so the way
in which the material is explained to the two readerships must be based on your
judgement, in each case, of what is common knowledge and what is unfamiliar. The
nature of the audience may even determine what can be reported.
Making choices about the content of a paper places limits on its scope; these
choices identify material to be excluded. Broadly speaking, many research pro-
grams are a cycle of innovation and evaluation, with the answers or resolution of
one investigation creating the questions that lead to the next. An advance in, say,
string sorting might well have implications for integer sorting, and further work
could pursue these implications. But at some point it is necessary to stop under-
taking new work and write up what has been achieved so far. The new ideas may
well be exciting—and less stale than the work that has been preoccupying you for
months—but they are likely to be less well understood, and completing the old work
is more important than trying to include too many results. If the newer work can be
published independently, then write it up separately. A long, complex paper, how-
ever big a breakthrough it represents, is hard to referee. From an editor's perspective,
accepting such a paper may be difficult to justify if it squeezes out several other
contributions.
Another element in the process of developing a paper is deciding where the work
might be published. There are many factors that should be considered when making
this decision, such as relevance to your topic and how your work measures against
the standard for that forum. In particular, the venue partly determines the scope of
a paper. For example, is there a page limit? Are there specific conventions to be
observed? Are the other papers in that venue primarily theoretical or experimental?
What prior knowledge or background is a reader likely to have? Do the editors require
that your code be available online? If you select a particular forum but haven't cited
any papers that have appeared there, you may have made the wrong choice.
Once the material for a paper has been collected it has to be organized into a
coherent self-contained narrative, which ultimately will form the body of the write-
up. Turning this narrative into awrite-up involves putting it in the formof an academic
paper: including an introduction, a bibliography, and so on. These issues are discussed
later in this chapter.
