Information Technology Reference
In-Depth Information
The body should be reasonably independent of other papers. If, to understand
your paper, the reader must find specialized literature such as your earlier papers or
an obscure paper by your advisor, then its audience will be limited.
In some disciplines, research papers have highly standardized structures. Editors
may require, for example, that you use only the four headings Introduction-Methods-
Results-Discussion. This convention has not taken hold in computer science, and in
some cases such a structure impedes a clear explanation of the work. For example,
use of fixed headings may prohibit development of a complex explanation in stages.
In work combining two query resolution techniques, we had to determine how they
would interact, based on a fresh evaluation of how they behaved independently.
The final structure was, in effect, Introduction-Background-Methods-Results-Dis-
cussion-Methods-Results-Discussion.
Even if the standardized section names are not used, the body needs these elements,
if not necessarily under their standard headings. Components of the body might
include, among other things, background, previous work, proposals, experimental
design, analysis, results, and discussion. Specific research projects suggest specific
headings. For the “compression for fast external sorting” project sketched earlier,
the complete set of section headings might be:
1. Introduction
2. External sorting
3. Compression techniques for database systems
4. Sorting with compression
5. Experimental setup
6. Results and discussion
7. Conclusions
The wording of these headings does not follow the standard form, but the intent of
the wording is the same. Sections 2 and 3 are the background; Section4 contains
novel algorithms, and Sections 4 and 5 together are the methods.
The background material can be entirely separate from the discussion of pre-
vious work on the same problem. The former is the knowledge the reader needs
to understand your contribution. The latter is, often, alternative approaches that are
superseded by your work. Together, the discussion of background and previous work
also introduce the state of the art and its failings, the importance and circumstances
of the research question, and benchmarks or baselines that the new work should be
compared to.
A body that consists of descriptions of algorithms followed by a dump of unin-
terpreted experimental results is not sound science. In such a paper, the context of
prior work is not explained, as readers are left to draw their own inferences about
what the results mean.
The results section is an assembly of evidence on which the key arguments are
based. This typically includes presentations of experimental outcomes, theorems,
proofs, analyses of data, and tabulations of investigative outcomes and discoveries.
The arguments then convey what the results mean—that is, they need to be explained,
analyzed, and interpreted. For the data, readers need to know how the data was
Search WWH ::




Custom Search