4

I am a PhD student in computer science (theory). I am currently writing a research paper. It is a short paper consisting of about 7-8 pages. I have written it two time and have taken few advice from research supervisor. I am struggling with the part where we give the overview of our approach.

Problems I am facing :

  1. I can not write much in this part as it is just an overview.
  2. I have to be consistent with the symbol's I used in the main part of research paper.
  3. How much should I give in the overview section?
  4. Should I try to avoid symbols, proofs etc. in the overview section?

Question: How to convey your main idea in a research paper?

4

One of the important and difficult skills to learn in technical presentation is how to tell the same story at multiple levels of detail, from a single sentence to a multi-hour tutorial or long-form journal paper.

Moreover, any well-organized presentation (written or spoken) should typically present at multiple levels of resolution. As one of my grad school mentors used to say: "Tell them what you're going to tell them, then tell it to them, then tell them what you told them." This is not redundancy, but rather priming the reader to understand how the pieces fit together and summarizing to help them retain it.

In most good papers, I would thus expect to see the "main idea" presented either five or six times, at various levels of resolution:

  1. As a few words in the title, naming the most important "keyword" ideas.
  2. As a few sentences in the abstract, identifying a problem, giving an approach, and claiming evidence of progress.
  3. As one or two paragraphs in the introduction, sketching the key ideas and how the rest of the paper is organized.
  4. Optionally, as an overview section providing a sketch of the main sections and the grounding of the approach in prior work.
  5. In full detail in the main sections.
  6. In a few summarizing sentences in the final section, highlighting the most important things to remember from the main sections.

So how do you write at a "lower resolution"? I recommend thinking about the difference between what you are doing and how you are doing it. For example, the statement of a theorem is a "what" and the proof is a "how". In a lower resolution, however, the "what" can become a how, presented consistently but less formally. For example, that theorem may be one of several things that you prove as the "how" to get to the "what" of collectively proving a key result. One can then continue going to lower and lower resolutions. For example, all the work on proofs may be one "how" that combines with an implementation "how" and an experiment "how" to get to the "what" of demonstrating progress against the problem you are trying to address.

In short: remain consistent in your terminology, but drop as many levels of detail as you need to fit the space available for each telling of the main idea.

  • First part is good. But you lost me after 'So how do you write...'. – Prof. Santa Claus Aug 11 '18 at 10:45
  • I recommend thinking about the difference between what you are doing and how you are doing it — To put this in more familiar terms, since you’re a theoretical computer scientist: Think about the difference between specifying the problems you’re solving and describing the algorithms you’re using to solve them. Equivalently, organize your results in breadth-first order, not depth-first order. – JeffE Aug 12 '18 at 3:26
2

Many great researchers share their tips of writing papers, here are my favorites:

All so read the papers that gets the awards, and papers of authors who are known to spend great effort on polishing their papers.

Back to your questions:

  1. I can not write much in this part as it is just an overview.

It's difficult to say without knowing what kind of paper you are writing. In my field, even theoretical papers are often illustrated with a motivating example. This is a paper with a very good Overview section (best paper award in PLDI 2013).

  1. I have to be consistent with the symbol's I used in the main part of research paper.

A trick that most Latex users know: don't use a symbol directly, e.g. \Sigma, instead define a new command, e.g. \newcommand{\mySymbol}{\Sigma}, and use \mySymbol everywhere to be consistent. It is also very easy if you decide later that you want to replace \Sigma with a different symbol.

  1. How much should I give in the overview section?

Around 1 page for a papers with 7 - 8 pages.

  1. Should I try to avoid symbols, proofs etc. in the overview section?

Yes, as much as you can. When this is not possible, explain them clearly and say something such as "formal definition will be provided in subsequent section", etc

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.