Writing Technical Articles

Read Strunk and White, Elements of Style. Again.

Give the paper to somebody else to read. If you can, find two people: one person familiar with the technical matter, another only generally familiar with the area.

Papers can be divided roughly into two categories, namely original research papers and survey papers. There are papers that combine the two elements, but most publication venues either only accept one or the other type or require the author to identify whether the paper should be evaluated as a research contribution or a survey paper. (Most research papers contain a "related work" section that can be considered a survey, but it is usually brief compared to the rest of the paper.)

Research Papers

A good research paper has a clear statement of the problem the paper is addressing, the proposed solution(s), and results achieved. It describes clearly what has been done before on the problem, and what is new.

The goal of a paper is to describe novel technical results. There are four types of technical results:

1. An algorithm;
2. A system construct: such as hardware design, software system, protocol, etc.;

   One goal of the paper is to ensure that the next person who designs a system like yours doesn't make the same mistakes and takes advantage of some of your best solutions. So make sure that the hard problems (and their solutions) are discussed and the non-obvious mistakes (and how to avoid them) are discussed. (Craig Partridge)

3. A performance evaluation: obtained through analyses, simulation or measurements;
4. A theory: consisting of a collection of theorems.

• describing the results in sufficient details to establish their validity;
• identifying the novel aspects of the results, i.e., what new knowledge is reported and what makes it non-obvious;
• identifying the significance of the results: what improvements and impact do they suggest.

Paper Structure

• Typical outline of a paper is:
  • Abstract, typically not more than 100-150 words;
  • Introduction (brief!): introduce problem, outline solution; the statement of the problem should include a clear statement why the problem is important (or interesting).
  • Related Work (or before summary). Hint: In the case of a conference, make sure to cite the work of the PC co-chairs and as many other PC members as are remotely plausible, as well as from anything relevant from the previous two proceedings. In the case of a journal or magazine, cite anything relevant from last 2-3 years or so volumes.
  • Outline of the rest of the paper: "The remainder of the paper is organized as follows. In Section 2, we introduce ..Section 3 describes ... Finally, we describe future work in Section 5." [Note that Section is capitalized. Also, vary expression between section as active part of sentence to "in Section, we".]
  • Body of paper
    • problem
    • approach, architecture
    • results

    The body should contain sufficient motivation, with at least one example scenario, preferably two, with illustrating figures, followed by a crisp generic problem statement model, i.e., functionality, particularly emphasizing "new" functionality. The paper may or may not include formalisms, O(n) kinds of characterizations go here, not in evaluation.

    Architecture of proposed system(s) to achieve this model should be more generic than your own peculiar implementation. Always include at least one figure.

    Realization: contains actual implementation details when implementing architecture isn't totally straightforward. Mention briefly implementation language, platform, location, dependencies on other packages and minimum resource usage if pertinent.

    Evaluation: How does it really work in practice? Provide real or simulated performance metrics, end-user studies, mention external technology adoptors if any, etc.

  • Related work, if not done at the beginning
  • Summary and Future Work
    • often repeats the main result
  • Acknowledgements
  • Bibliography
  • Appendix (to be cut first if forced to):
    • detailed protocol descriptions
    • proofs with more than two lines
    • other low-level but important details

It is recommended that you write the approach and results sections first, which go together. Then problem section, if it is separate from the introduction. Then the conclusions, then the intro. Write the intro last since it glosses the conclusions in one of the last paragraphs. Finally, write the abstract. Last, give your paper a title.

Abstract

• The abstract must not contain references, as it may be used without the main article. It is acceptable, although not common, to identify work by author, abbreviation or RFC number. (For example, "Our algorithm is based upon the work by Smith and Wesson.")
• Avoid use of "in this paper" in the abstract. What other paper would you be talking about here?
• Avoid general motivation in the abstract. You do not have to justify the importance of the Internet or explain what QoS is.
• Highlight not just the problem, but also the principal results. Many people read abstracts and then decide whether to bother with the rest of the paper.
• Since the abstract will be used by search engines, be sure that terms that identify your work are found there. In particular, the name of any protocol or system developed and the general area ("quality of service", "protocol verification", "service creation environment") should be contained in the abstract.
• Avoid equations and math. Exceptions: Your paper proposes E = m c ².

Introduction

• Avoid stock and cliche phrases such as "recent advances in XYZ" or anything alluding to the growth of the Internet.
• Be sure that the introduction lets the reader know what this paper is about, not just how important your general area of research is. Readers won't stick with you for three pages to find out what you are talking about.
• The introduction must motivate your work by pinpointing the problem you are addressing and then give an overview of your approach and/or contributions (and perhaps even a general description of your results). In this way, the intro sets up my expectations for the rest of your paper -- it provides the context, and a preview.
• Repeating the abstract in the introduction is a waste of space.
• Example bad introduction:

  Here at the institute for computer research, me and my colleagues have created the SUPERGP system and have applied it to several toy problems. We had previously fumbled with earlier versions of SUPERGPSYSTEM for a while. This system allows the programmer to easily try lots of parameters, and problems, but incorporates a special constraint system for parameter settings and LISP S-expression parenthesis counting.

  The search space of GP is large and many things we are thinking about putting into the supergpsystem will make this space much more colorful.

• A pretty good introduction, drawn from Eric Siegel's class:

  Many new domains for genetic programming require evolved programs to be executed for longer amounts of time. For example, it is beneficial to give evolved programs direct access to low-level data arrays, as in some approaches to signal processing \cite{teller5}, and protein segment classification \cite{handley,koza6}. This type of system automatically performs more problem-specific engineering than a system that accesses highly preprocessed data. However, evolved programs may require more time to execute, since they are solving a harder task.

  (PREVIOUS OR OBVIOUS APPROACH: (Note that you can also have a related work section that gives more details about previous work.)) One way to control the execution time of evolved programs is to impose an absolute time limit. However, this is too constraining if some test cases require more processing time than others. To use computation time efficiently, evolved programs must take extra time when it is necessary to perform well, but also spend less time whenever possible.

  (APPROACH/SOLUTION/CONTRIBUTION. THE *FIRST* SENTENCE OF A PARAGRAPH LIKE THIS SHOULD SAY WHAT THE CONTRIBUTION IS. ALSO GLOSS THE RESULTS.) In this chapter, we introduce a method that gives evolved programs the incentive to strategically allocate computation time among fitness cases. Specifically, with an {\it aggregate computation time ceiling} imposed over a series of fitness cases, evolved programs dynamically choose when to stop processing each fitness case. We present experiments that show that programs evolved using this form of fitness take less time per test case on average, with minimal damage to domain performance. We also discuss the implications of such a time constraint, as well as its differences from other approaches to {\it multiobjective problems}. The dynamic use of resources other than computation time, e.g. memory or fuel, may also result from placing an aggregate limit over a series of fitness cases.

  (OVERVIEW:) The following section surveys related work in both optimizing the execution time of evolved programs and evolution over Turing-complete representations. Next we introduce the game Tetris as a test problem. This is followed by a description of the aggregate computation time ceiling, and its application to Tetris in particular. We then present experimental results, discuss other current efforts with Tetris, and end with conclusions and future work.

Body of Paper

• Avoid use of passive tense if at all possible. Example: "In each reservation request message, a refresh interval used by the sender is included." reads better and shorter as "Each ... message includes ..."
• Use strong verbs instead of lots of nouns. Examples:

  verbose, weak verbs, bad	short, strong, good
  make assumption	assume
  is a function of	depends on
  is an illustration	illustrates
  is a requirement	requires, need to

• Check for missing articles, particularly if your native tongue doesn't have them. Roughly, concepts and classes of things don't, most everything else more specific does. ("Routers route packets. The router architecture we consider uses small rodents.") [NEED POINTER HERE]
• Use consistent tense (present, usually, unless reporting results achieved in earlier papers).
• Use hyphens for concatenated words: "end-to-end architecture", "real-time operating system", "per-flow queueing", ...
• Numbers ten or less are spelled out: "It consists of three fields", not "3 fields".
• Avoid in-line enumeration like: "Packets can be (a) lost, (b) stolen, (c) get wet." The enumeration only interrupts the flow of thought.
• Avoid itemization (bullets), as they take up extra space and make the paper read like PowerPoint slides. Bullets can be used effectively for emphasis of key points. If you want to describe components or algorithms, often the description environment works better, as it highlights the term, providing a low-level section delineation.
• Instead of "[1] shows" use "Smith [1] showed" or "Smith and Jones [1] showed" or "Smith et al [1] showed" (if more than two authors). Or, alternatively, "the foobar protocol [1] is an example ...". This keeps the reader from having to flip back to the references, as they'll recognize many citations by either author name or project name. No need to refer to RFC numbers in the text (except in RFCs and Internet Drafts). Exception for very low-level presentation: "RFC822-style addresses".
• Use normal capitalization in captions ("This is a caption", not "This is a Caption").
• Avoid excessive parenthesized remarks as they make the text hard to read; fold into the main sentence. Check whether the publication allows footnotes - some magazines frown upon them. More than two footnotes per page is a bad sign. You probably should have applied to law school instead.
• There is no space between the text and the superscript for the footnote. I.e., it's text\footnote{} rather than text \footnote{}.
• Check that abbreviations are always explained before use. Exceptions, when addressed to the appropriate networking audience: ATM, BGP, ftp, HTTP, IP, IPv6, RSVP, TCP, UDP, RTP, RIP, OSPF, BGP, SS7. Be particularly aware of the net-head, bell-head perspective. Even basic terms like PSTN and POTS aren't taught to CS students... For other audiences, even terms like ATM are worth expanding, as your reader might wonder why ATM has anything to do with cells rather than little green pieces of paper.
• Avoid capitalization of terms. Your paper is not the U.S. Constitution or Declaration of Independence. Technical terms are in lower-case, although some people use upper case when explaining an acronym, as in "Asynchronous Transfer Mode (ATM)".
• Each paragraph should have a lead sentence summarizing its content. If this doesn't work naturally, the paragraph is probably too short. Try reading just the first lines of each paragraph - the paper should still make sense. For example,

  There are two service models, integrated and differentiated service. Integrated service follows the German approach that anything that isn't explicitly allowed is verboten. It strictly regulates traffic, but also makes the trains run on time. Differentiated service follows the Animal Farm appraoch, where some traffic is more equal than others. It seems simpler, until one has to worry about proletariat traffic dressing up as the aristocracy.

• $i$th, not $i-th$.
• Units are always in roman font, never italics or math mode. Units are set off by one (thin) space from the number. In LaTeX, use ~ to avoid splitting number and units across two lines. \; for a thin space.
• Use \, in LaTeX between the number and the unit, as it inserts the appropriate thin space. A regular white space often looks too wide.
• Use "kb/s" or "Mb/s", not "kbps" or "Mbps" - the latter are not scientific units. Be careful to distinguish "Mb" (Megabit) and "MB" (Megabytes), in particular "kb" (1000 bits) and "KB" (1024 bytes).
• Use "ms", not "msec", for milliseconds.
• There's no need to enclose numbers in $$ (math mode).
• Avoid "etc."; use "for example", "such as", "among others" or, better yet, try to give a complete list (unless citing, for example, a list of products known to be incomplete), even if abstract. See also Strunk and White:

  Etc.: Not to be used of persons. Equivalent to and the rest, and so forth, and hence not to be used if one of these would be insufficient, that is, if the reader would be left in doubt as to any important particulars. Least open to objection when it represents the last terms of a list already given in full, or immaterial words at the end of a quotation. At the end of a list introduced by such as, for example, or any similar expression, etc. is incorrect.

• Avoid bulleted lists of one-sentence paragraphs. They make your paper look like a slide presentation and interfere with smooth reading.
• Avoid excessive use of "i.e.". Vary your expression: "such as", "this means that", "because", .... "I.e." is not the universal conjunction!
• Remember that "i.e." and "e.g." are always followed by a comma.
• Never use "related works" unless you are talking about works of art. It's "related work".
• Use "in Figure 1" instead of "following figure" since figures may get moved during the publication process. Don't assume that the LaTeX figure stays where you put it.
• Do not use GIF images for figures, as GIFs produce horrible print quality and are huge. Export into PostScript. At that stage, you'll learn to "appreciate" Microsoft products. xfig and gnuplot generally produce PostScript that can be included without difficulties.
• Section, Figure and Table are capitalized, as in "As discussed in Section 3". Figure can be abbreviated, but the others are not usually, but that's a matter of taste - just be consistent.
• Figures show, depict, indicate, illustrate. Avoid "(refer to Fig. 17)". Often, it is enough to simply put the figure reference in parenthesis: "Packet droppers (Fig. 17) have a pipe to the bit bucket, which is emptied every night."
• Use \cite{a,b,c}, not \cite{a} \cite{b} \cite{c}.
• Use the \usepackage{times} option for LaTeX2e - it comes out much nicer on printers with different resolutions. Plus, compared to cmr, it probably squeezes an extra 10% of text out of your conference allotment.
• For uniformity, use the LaTeX2e graphics set, not the earlier psfigure set:

  \usepackage{graphics}
  ...
  \begin{figure}
  \resizebox{!}{0.5\textheight}{\includegraphics{foo.eps}}
  \caption{Some figure}
  \label{fig:figure}
  \end{figure}
  

Bibliography

• Avoid use of et al. in a bibliography unless list is very long (five or more authors). The author subsumed into et al. may be your advisor or the reviewer... Note punctuation of et al..
• If writing about networks or multimedia, use the network bibliography. All entries not found there should be sent to me. A listing of frequently-used references for networks is available.
• Internet drafts must be marked ``work in progress''.
• Book citations include publication years, but no ISBN number.
• It is now acceptable to include URLs to material, but it is probably bad form to include a URL pointing to the author's web page for papers published in IEEE and ACM publications, given the copyright situation. Use it for software and other non-library material. Avoid long URLs; it may be sufficient to point to the general page and let the reader find the material. General URLs are also less likely to change.
• Leave a space between first names and last name, i.e., "J. P. Doe", not "J.P.Doe".

Acknowledgements

• Generally, anonymous reviewers don't get acknowledged, unless they really provided an exceptional level of feedback or insight. Rather than "We thank X for helping us with Y", you might vary this as "X helped with Y.".

Things to Avoid

• Too much motivational material: 3 reasons are enough -- and they should be described very briefly.
• Describing the obvious parts of the result: "Obvious" is defined as any result that a graduate of our program would suggest as a solution if you pose the problem that the result solves.
• Describing unnecessary details: a detail is unnecessary, if its omission will not harm the reader's ability to understand the important novel aspects of the result.

Guidelines for Experimental Papers

1. Papers that introduce a new learning "setting" or type of application should justify the relevance and importance of this setting, for example, based on its utility in applications, its appropriateness as a model of human or animal learning, or its importance in addressing fundamental questions in machine learning.
2. Papers describing a new algorithm should be clear, precise, and written in a way that allows the reader to compare the algorithm to other algorithms. For example, most learning algorithms can be viewed as optimizing (at least approximately) some measure of performance. A good way to describe a new algorithm is to make this performance measure explicit. Another useful way of describing an algorithm is to define the space of hypotheses that it searches when optimizing the performance measure.
3. Papers introducing a new algorithm should conduct experiments comparing it to state-of-the-art algorithms for the same or similar problems. Where possible, performance should also be compared against an absolute standard of ideal performance. Performance should also be compared against a naive standard (e.g., random guessing, guessing the most common class, etc.) as well. Unusual performance criteria should be carefully defined and justified.
4. All experiments must include measures of uncertainty of the conclusions. These typically take the form of confidence intervals, statistical tests, or estimates of standard error. Proper experimental methodology should be employed. For example, if "test sets" are used to measure generalization performance, no information from the test set should be available to the learning process.
5. Descriptions of the software and data sufficient to replicate the experiments must be included in the paper. Once the paper has appeared in Machine Learning, authors are strongly urged to make the data used in experiments available to other scientists wishing to replicate the experiments. An excellent way to achieve this is to deposit the data sets at the Irvine Repository of Machine Learning Databases. Another good option is to add your data sets to the DELVE benchmark collection at the University of Toronto. For proprietary data sets, authors are encouraged to develop synthetic data sets having the same statistical properties. These synthetic data sets can then be made freely available.
6. Conclusions drawn from a series of experimental runs should be clearly stated. Graphical display of experimental data can be very effective. Supporting tables of exact numerical results from experiments should be provided in an appendix.
7. Limitations of the algorithm should be described in detail. Interesting cases where an algorithm fails are important in clarifying the range of applicability of an algorithm.

The Conference Review Process

It is hard to generalize the review process for conferences, but most reputable conferences operate according to these basic rules:

1. The paper is submitted to the technical program chair(s).
2. The TPchair assigns the paper to one or more technical program committee members, hopefully experts in their field. The identity of this TPC member is kept secret.
3. The TPC member usually provides a review, but may also be asked to find between one and three reviewers who are not members of the TPC. They may be colleagues of the reviewer at the same institution, his or her graduate students or somebody listed in the references. The graduate student reviews can be quite helpful, since these reviewers often provide more detailed criticism rather than blanket dismissal. Any good conference will strive to provide at least three reviews, however, since conferences operate under tight deadlines and not all reviewers deliver as promised, it is not uncommon that you receive only two reviews.
4. The TPChair then collects the reviews and sorts the papers according to their average review scores.
5. The TPC, or usuaally a subset that can make the meeting, then meets Usually, the bottom third and the top third are rejected and accepted without (much) further discussion. The papers discussed are those in the middle of the range, where a TPC member feels strongly that the paper ended up in the wrong bin, or where the review scores differ significantly, in particular if there are only two reviews.

Other References

• Oded Goldreich wrote an essay entitled "How not to write a paper", with recommendations on style and organization.
• Don Knuth has online the TeX source of a book on "Mathematical Writing" (also useful for Computer Science).
• The Elements of Style. William Strunk Jr. and E.B. White. Macmillan Publishing Co., New York, 1979.

  This is an amazing little book that you can read in a few hours. Your writing style will never be the same afterwards. This $8 book is the best investment you can ever make.

• Bugs in Writing. Lyn Dupre', (2nd ed.)

  This is a great book that expands on Strunk&White. It has more examples, and was written by an author who edited numerous technical publications, many of which were in computer science.

• The Chicago Manual of Style, Univ. of Chicago Press.

  This is the bible for American academic style. It's long and heavy, but has everything you ever want to know about style. When in doubt, or if you get conflicting stylistic advice, following The Chicago Manual of Style is your best choice.

• A Handbook for Scholars by Mary Claire van Leunen; Alfred Knopf, Publisher.

  This is another useful book written for publishing (computer) scientists.

• The UIST Guide for Authors is geared towards a specific conference, but the general process and guidelines are similar to many other conferences.
• The Science of Scientific Writing, George D. Gopen and Judith A. Swan, In American Scientist, Vol. 78, No. 6 (Nov-Dec, 1990), pp. 550-558.

  This is a useful article that teaches scientists how to write single sentences and paragraphs.

• Roy Levin and David D. Redell, An evaluation of the ninth SOSP submissions -or- How (and how not) to write a good systems paper, ACM SIGOPS Operating Systems Review 17(3):35-40 (July, 1983).
• Alan Snyder, How to get your paper accepted at OOPSLA, OOPSLA '91 Proceedings, pp. 359-363.
• Ralph Johnson et al, How to get a paper accepted at OOPSLA, Panel at OOPSLA'93, pp 429-436.
• Craig Partridge, How to Increase the Chances Your Paper is Accepted at ACM SIGCOMML.

  Generally useful advice that also applies to other networking conferences.

• What kinds of papers does USENIX publish?
• Alan Jay Smith, The Task of the Referee, IEEE Computer 23(4):65-71 (April 1990).

Contributors