University of Virginia Department of
    Computer Science

The Graduate Student's Guide to Writing Scientific Papers

Motivation

Becoming a good writer is perhaps the single greatest thing you can do to further your research career. Your ideas and engineering, algorithmic, or theoretical solutions can be strong and original, but if you communicate your ideas poorly then you will diminish the impact of your ideas and solutions. A cogent and concise description of your research helps readers to understand the core issues which you are dealing with, rather than getting bogged down in the details of their communication.

It is sometimes helpful to think of your paper as the user interface to your research. If it is confusing and poorly presented then nobody will take note of it. By contrast, a clear, fluid, compact, and visually inviting presentation will entice your audience to

Executive summary: good writing is your ticket to fame, glory and funding.

Understand that the process of writing almost always leads you to a more thorough comprehension of your topic. You should never consider the process of writing as an extraneous task that prevents you from doing the very work which you are trying to describe. If you want to describe a theory or algorithm well, you must understand it completely. By seeking to cogently describe your work, you will come to understand it better, and thus will improve on the work itself.

The bulleted list is your friend!

The bulleted list is a compact, effective, and eye-catching way to get lots of information across to your reader in a hurry.

Use bulleted lists for 3 or more items. Sometimes bulleted lists are appropriate for 2 items, but if so set it up such that the reader expects only 2 items. A bullet should almost never be used for just 1 item, unless you are really trying to call attention to it ... and in most cases you'd want a label instead (e.g. Thm #1: or whatever).

If you can summarize the bullet in just one or two words, do so and bold or italic those words.

Be concise

First, you should admit to yourself that almost nobody is going to read your whole paper. Most will just read the abstract. This means that the more good stuff you can cram into these 200 words or so, the more you will be able to convey about your work. Heck, you may even be able to connive someone to read your introduction as well. In fact, you should make certain that your main idea / contribution is right there in the first sentence of the abstract.

Use no unnecessary words. As you read your draft, always ask yourself, "could I convey the same ideas with fewer words?" From an engineering standpoint, you want the text of your document to have a high signal to noise ratio.

Keep paragraphs fairly short. Look to break up long paragraphs. If the paragraph is lengthy b/c it addresses a series of points, break it up using a bulleted list.

Good writing is knowing bad writing when you see it, and fixing it

A good writer is someone who is good at seeing where the mistakes are, and fixing them. Learn to find chunks of text that don't meld well. Have people read your draft. Places where they suggest edits, or ask questions, are likely tripping points. Focus on these first.

Read Marc Raibert's paper.

For scientific papers, pay particular attention to his points about "spilling the beans." Don't make your paper a mystery: divulge everything from the start. If you have good stuff that someone should be interested, put it right up there in the abstract and introduction, to try to hook your reader. Put your most important idea(s) right into the title.

Summaries, Conclusions, and Future Work sections

Avoid summary/conclusion sections that merely reiterate the points that you have raised previously. Your central message might be retreated, but don't give an exhaustive revisit of your paper. In particular, don't paste sections of text or sentences the reader has already seen into your conclusion. At worst, the reader may feel insulted; at best, the reader will stop reading the paper or will just skim the rest, since they've already seen everything you're telling them again.

The conclusion is a great place for bulleted lists that quickly (in a few words per bullet) summarize your main ideas or, for example, the main advantages of your technique.

Also use your conclusion as a point to jump off into your future work section. You might point out some of the shortcomings of the work (if you haven't already done so) and then quickly (in the future work section) follow with how you plan to fix or ameliorate the problems you encountered.

For future work, be sure to suggest some fairly trivial extensions that you already are planning or even starting to implement. Use these to lead in to your more radical or futuristic extensions.

Writers often like to end their scientific papers on visionary note. be extremely careful if you do this. It is okay to hint at where your work might go in say, 2 or 3 years. It is also okay to suggest where your work could go if a tractable technological impediment were removed add concrete examples . But do not turn into a science fiction geek here. Don't suggest there's a pot of gold if you know darn well that it probably doesn't exist.

Eliminate Weasel words

simple (simply), very, quite, obvious, basic (basically), almost, ...

With rare exceptions, these words will weaken your text. Find all occurrences and eradicate them, even if it means recasting your sentences a bit.

Be concrete

Don't be too abstract, even when you are explaining abstract concepts. Be sure to give concrete examples of what you mean...

Write using verbs and nouns. In scientific papers it is especially important to Avoid unnecessary adjectives and flowery prose. This is one of the most difficult guidelines to apply... it will take practice. Give examples

Read Strunk and White.

Don't open yourself to attack by the reviewers.

Unless you are absolutely certain you can defend yourself against all attacks, do not make overly definitive statements, unless they specifically describe your work:

nobody has tried	the research we are aware of has not
there is no		existing evidence suggests that...
etc...

Better Examples

Don't flame out the work of your colleagues. Just politely explain (perhaps using a bulleted list) the difference between your approaches and design tradeoffs. Be sure to mention what your colleague's work handles well. Your colleague may very well end up reviewing the paper, so be diplomatic. And, if you have made a mistake, you at least have a path to escape...

Use Lots Of Figures And Photographs

Use figures and photographs wherever possible. You can frequently express ideas that would take whole paragraphs with using one simple figure. If a picture can be used to help explain your main idea, (1) be sure to make a good figure and include it, and (2) put it right on the front page of your work, right below the abstract or in the introduction. The picture will become a sort of symbol for your paper that will help people remember it, and the picture itself will attract attention.

If you're clever, figures and photographs are a good way to provide a "digest" of your paper. Many people will read only the abstract and then flip through, looking at the just the pictures. If the pictures provide a high-level overview of the paper, you may draw the reader's interest and get him to read the whole paper, or at the very least, you will succeed in communicating your main idea to the reader.

A Few Graphic Design Guidelines

find someone with artistic skill or taste and get them to comment on your figures. If you're really lucky, there will be such a person to help you draw such figures: except for graphs / charts, a photograph or hand-drawn image almost always looks better than a computer drawn image (some modern paint programs excepted)

This is even more true for slide presentations of your work...

Iterative Design

Start with bad writing. Keep iterating until it is good writing.

A very gifted writer might be able to write a decent paper with as few as 6 iterations. A skilled English major could probably get by with about 12 or so. For the rest of us mere mortals, and graduate engineering students in particular, 20 or more iterations of the document would not be an unreasonable number.

If you are out of practice, set aside 2-3 weeks to work on your paper, and try to iterate on the design of the paper almost every day (even if you just re-organize a couple of paragraphs). You will find that not only does the writing improve, but the scientific merits of the work that you are describing will improve as well.

To explain something clearly requires that you understand it thoroughly. The process of writing the document will help you to discover errors and weak spots in your work. By setting aside lots of time to write the paper, you are also allowing yourself time to patch up the weak areas in your work.

Appeal

Like a good film, your paper needs to have appeal. You've got to hook the reader in and keep them interested. The paper has to have a certain sense of momentum. "phrasing" borrow techniques from theatre/film. visuals are one example.



HelpNet
HelpNet was created by the 1995 incoming graduate class. It is only occasionally updated.
Department of Computer Science
School of Engineering, University of Virginia
151 Engineer's Way, P.O. Box 400740
Charlottesville, Virginia 22904-4740

(434) 982-2200  Fax: (434) 982-2214
Web Comments: webteam@cs.virginia.edu
Admissions Inquiries: inquiry@cs.virginia.edu
Site directory, Other addresses
Server statistics
© Created by the CS Web Team