Write It Right
by Judy Goldsmith and Robert H. Sloan
- Provide a summary or abstract
- Give enough information that a reader can follow your claims.
- Reference the key papers on the subject, though longer discussions are often postponed to the journal version.
It is sometimes useful to make a technical report version, with more information, available online.
Review process
Conference reviewing standards tend to focus more on novelty, applicability, and plausibility rather than correctness. This is not an excuse to be sloppy, but a warning to focus on making clear to the reader what is important and new about your work.
Conference reviewers tend to have a lot of papers to review in a too-short time. Make your major points clearly and early. Write carefully and well. When all else is equal, reviewers prefer well-written papers to those that make them hunt for definitions or guess about notation, meaning, or importance.
Journal Papers
Journal papers are "archival, " meaning that, in the days before the World Wide Web, they persisted in libraries and were indexed in publications, such as the Science Citation Index, which used to be published annually in many thin-paged, small-print volumes.
Intended audience
It is still assumed that journal papers will outlast conference papers or technical reports, albeit perhaps not in bound volumes on bookshelves. Thus, you are writing both for current experts in your field, and anyone who is now, or may someday be, interested in your work.
You cannot assume that they read current blogs or papers or that they have gone to the same conferences you have. Your article should be self-sufficient. Consider that you may be ahead of your time--your new result may be only moderately interesting today, but may be the crucial ingredient for some important research in a decade or two--if your paper is written so that the researcher who needs your result in 20 years' time can read it.
Of course, as with all scholarly writing, a journal paper should explain the importance of the problem considered and the significance of the solution.
Content
Journal papers have the space, generally, to give sufficient technical detail for a reader to verify your results. There is also space to seriously consider related work.
It is useful to end a paper with open questions. This is a service to young researchers, and has the potential to get your work cited by those who attempt to solve the open problems. However, it is a bad idea to list problems on which you are currently actively working, because you risk being "scooped. "
Review Process
Depending on the journal, a paper can be reviewed once, twice, or more times. Reviewers are selected for their expertise, and are expected to verify results as well as judge whether the material is of interest, and relevant to the scope of the journal. They will not repeat experiments, but will judge whether the experiment and results appear valid. They are expected to comment on relevance, clarity, appropriateness, and correctness.
If your work involves theorem-proof type mathematics, be aware that software engineering journals vary on the standard of review for mathematical proofs. Assume the worst case--that any embarrassing bug in your proof that you would like a referee to miss will be caught, but that if you are relying on referees as the final set of verifiers, those verifiers will fail you.
The journal editor gathers one to five reports (most often, three), makes a decision, and sends you the reports, a summary, and a decision. The decision can be a rejection, a request for rewriting and re-reviewing, or acceptance. Even acceptances are often predicated on your making minor changes.
Survey Papers
A survey paper provides a wide view of a problem and the solutions available for that problem. It is more than a list of solutions, however. It puts the subfield and the problem into a broader context, explaining the importance of the variety of solutions.
Intended audience
The expected reader of a survey paper is not an expert in the field. She may be an expert in a related subfield, someone new to the field, or may be a more casual reader.
Content
A good survey paper discusses the relative merits and drawbacks of the different solutions, and in particular, makes clear what variants of the problem each approach solves.
Review Process
Survey papers normally have the same review process as the journal in which they are published, which is usually a journal of the type described in the previous section.
Popular Articles
This category covers technical magazines published by the professional societies, such as IEEE Software, as well as magazines such as Wired or Byte that you might see at a newsstand. Many, but not all, of the articles in the very widely distributed magazines, IEEE Computer and Communications of the ACM, are in this category. Scientific American and Science News Weekly are also examples of technically focused, popular magazines. The aim of such articles is to increase interest in a field, problem, or publication.
Intended audience
The intended audience includes experts working in industry, as well as nonexperts.
Content
In software engineering, the most significant difference between "popular " and "scholarly " articles is the degree of research required. For example, to publish an article on requirements engineering in the International Conference on Requirements Engineering, one might conduct a field study of the success of a few different methods in practice.
On the other hand, a popular article on requirements engineering might give readers advice about what appears to be best practice based on the author's reading of a few such field studies, and perhaps also her personal experience.
Review process
In the magazines published by the professional societies, submissions other than editorials and invited articles will go through the same review process as scholarly works. In all cases, your article will be reviewed for clarity, readability, interest, and correctness.
Dissertations
Dissertations are the culmination of graduate research in a particular field of study. They describe the student's original research, which should make a substantial and significant contribution to the field of study. Dissertations are required by most universities in order to receive a doctorate.
Intended audience
The dissertation may be used by others in the field as a reference, both for the subfield and for the author's work. Dissertations often make excellent starting places for new students.
Content
A dissertation is the one place where you are expected to give all the details of proofs and experiments. If you are not sure which graphs to include, you include all of them. You discuss the implications of your work, and you may speculate. You go into detail about related work and its relevance to your own work.
Review process
The student's committee is supposed to read the dissertation, and the advisor is expected to read it and respond with comments before the rest of the committee sees it. These readings, alas, do not always happen. When you read a dissertation, look for refereed publications based on the work in the dissertation.
Note that standards for master's theses vary from institution to institution.
Technical Reports
In the university environment, the purposes of technical reports are to timestamp your work in a competitive environment, to have something to show the boss while you await publication, and to electronically archive the details of experiments and proofs that do not fit into the publications so far.
In the business world, technical reports present data and conclusions about a particular topic. Examples include accident reports, environmental reclamation studies, clinical trial reports, design evaluations, and so on.
Intended audience
Anyone who wants an advanced peek at potential publications, or wants more details than current publications provide. The audience also might include stakeholders in the project or design being reported on. These stakeholders often, but not always, have a technical background in the subject matter.
Content
Technical reports can be about virtually any technical subject. Typically, they contain the following sections:
- Title page
- Table of contents
- Abstract
- Executive Summary
- Report Body (including graphics)
- Appendices
- Graphics that are large format (fold-out pages, engineering drawings, etc.)
- Glossary
- Index
Review process
For academic technical reports, there is no standard peer review process.
In business, technical reports often cover sensitive subjects (e.g., the Columbia shuttle accident) and are typically reviewed by subject matter experts, upper management, and sometimes legal or regulatory personnel.
White papers
In the university environment, white papers are pre-proposals.
In the business world, white papers typically tell customers and potential customers about a company's position on some aspect of the industry, explain how to use the company's products to better implement a best practice, or provide an explanation of a technical topic oriented toward a less technical audience. Many companies use white papers as a marketing tool.
Intended audience
In the university environment, white papers are pre-proposals, intended for a funding agency and possibly a small committee. They are not supposed to be circulated, except for review, except by the authors' choice.
In the business world, white papers typically are oriented toward a specific segment of customers, e.g., a conceptual architecture might be oriented toward software architects in the customer's organization.
Content
White papers need to state a problem and propose work toward a solution for that problem. If they are written in response to a call for white papers, they do not need to explain the problem in detail. They do need to give sufficient detail about the solution to make clear that it is new, and that you know enough to begin or continue work on the solution.
Review process
In the university environment, white papers are reviewed for relevance to the funder's mission, interest, and potential for success. They are not usually reviewed at a detailed level. Rather, they are considered for persuasiveness and plausibility.
In the business world, white papers typically provide specialized information or argue a point about a topic. They are typically reviewed by the project team or subject matter experts, as well as marketing, and sometimes regulatory.
Proposals
Most proposals are about work you expect to do, though successful proposals also usually refer to work you have already done. You must convince the readers that you are capable of doing the proposed work.
In the business world, proposals focus on showing the potential client that you understand their problem and have a cost-effective solution for it.
Intended audience
Most proposals are reviewed by a panel or a small group of researchers and/or developers, and by the funding agency/awarding company. You can assume in your readers a basic level of technical expertise, but cannot assume that the reviewers work in your subfield.
Content
You must persuade the audience that you can do the work requested by showing knowledge of the area, and if possible, by pointing to your own related publications, patents, or the like.
You need to describe the problem, provide a high-level synopsis of your approach, enough details to make your method clear, and an argument that what you are doing is better than what others have done. You should refer to evaluation criteria in the call for proposals, though you may suggest other criteria by which your work will be judged highly.
There are additional technical points in grant writing, including budgets and discussions of approval for any human-subjects research, but they are beyond the scope of this article. Most places where junior faculty or members of technical staff are writing their first grant proposals provide some support--do seek it out if you are a first-time grant writer.
Review process
In many funding agencies, a panel of experts reads the proposals and makes both individual and panel-wide recommendations to the funding agency. The agency then uses those recommendations to make a decision. Although panelists may make suggestions about details of the proposal, the proposal is usually not revised, though it might be rewritten and submitted elsewhere.
Next Column
The next column will discuss how to respond to peer-review comments.
References
ACM policy and procedures on plagiarism, October 2005.
Bovik, H. Q.; Goldsmith, J. Q.; Klapper, A. Q.; and M. Q. Littman. (April 2003) "Markov indecision processes: A formal model of decisionmaking under extreme confusion", Journal of Machine Learning Gossip, pages 1-9.
The Chicago Manual of Style. (2003) University of Chicago Press, 15th edition.
Dupr´e, L. (1998) BUGS in Writing, A Guide to Debugging Your Prose. Addison Wesley Professional.
Gibaldi, J. (2003) MLA Handbook for Writers of Research Papers. MLA Book Publications, 6th edition.
Gordon, K. E. (1993) The Deluxe Transitive Vampire: A Handbook of Grammar for the Innocent, the Eager and the Doomed. Pantheon.
Gordon, K. E. (2003) The New Well Tempered Sentence: A Punctuation Handbook for the Innocent, the Eager, and the Doomed. Mariner Books.
Handbook of Writing for the Mathematical Sciences. (1998) SIAM, second edition.
Johnson, J. (July 2005) On mathematical writing. Accessed on October 18, 2006.
Kitchenham, B.; Pfleeger, S. L.; Pickard, L.; Jones, P., Hoaglin, D.; Emam, K. E.; & Rosenberg, J. (2002) "Preliminary guidelines for empirical research in software engineering", IEEE Trans. Software Eng., 28(8):721-734.
Knuth, D. E.; Larrabee, T. L.; & Roberts, P. M. (1989) Mathematical Writing. Mathematical Association of America. Reprint (with corrections) of Technical Report 1193, Stanford University Computer Science Department, 1988.
Parberry, I. (1994) "A guide for new referees in theoretical computer science". Information and Computation, 112(1):96-116.
Shaw, M. (2003) Writing good software engineering research papers: Minitutorial. In Proc. 25th Int'l Conf. Software Eng. (ICSE 2003), pages 726-736.
Stone, H. S. (December 1992) "Copyrights and author responsibilities". IEEE Computer, pages 46-51.
Strunk, W. & White, E. B. (1999) The Elements of Style. Longman, 4th edition. The original 1918 edition of Strunk is available online at URL http://www.bartleby.com/141/.
Thomson Scientific. Science citation index. http://scientific.thomson.com/products/sci/
Voice of America. (23 August 2006) "Ohio University accuses engineering graduates of plagiarism". VOA News, August 2006. Downloaded October 11, 2006 from URL
http://www.voanews.com/specialenglish/archive/2006-08/2006-08-23-voa4.cfm.
Zobel, J. (2004) Writing for Computer Science. Springer, 2nd edition.
Judy Goldsmith is a computer science professor at the University of Kentucky. Her research interests include decision making under uncertainty; automation of information elicitation; preference elicitation, representation, and aggregation; computational learning theory, and structural complexity.
Robert H. Sloan is a professor (and acting department head) of computer science at the University of Illinois, Chicago. His research interests include application of computer science theory and algorithms to problems from artificial intelligence, especially machine learning ("computational learning theory") and knowledge representation;computer security, especially access control; and computer science education.
.