Gabriël Konat wrote a summary of the talk How to Write a Great Research Paper by Simon Peyton Jones
Whenever you have an idea, start writing a paper about it immediately, use the paper to do your research.
Why not write after doing research?:
Start writing (a paper) about ideas, no matter how simple, insignificant, or trivial the idea is. While you write the paper, the idea may develop and ramify. Something that looked boring can turn out to be rather interesting, but not always.
The greatest ideas are (literally) worthless if you keep them to yourself. A paper is not a mechanism for promotion, it is a mechanism for conveying ideas from your head to other people’s heads.
So you need to know what your (key) idea is!
Be explicit about what the key idea is. “The main idea of this paper is: …” or “In this section we present the main idea of this paper.” The reader should have no trouble figuring out what the main idea is, stating it explicitly ensures that they know.
Write a paper like explaining it to a colleague on a whiteboard. This explanation is usually in a much more accessible and engaging way.
Wherever the reader stops reading, they should take away something valuable, this narrative flow supports that. Also, every bit they read should make them want to read more.
Conference paper structure, and how many readers:
‘Each sentence, a couple of readers die off’
Introduction should be ONE PAGE!. Don’t put too much details of your idea in the introduction, because it will not fit.
Write the contributions first, because the list of contributions drives the entire paper. The main content of the paper substantiates the claims/contributions you have made.
Use a bulleted list of contributions with forward references to claims. Readers scan over the text, bullet points catch attention. Usually the first sentence of each bullet is read. The body of the paper provides evidence to support each claim. Evidence is something that makes the reader believe you are right, be it theorems, evaluation data or just text.
Using this style, readers can pick a contribution, go to the section that substantiates the claim and check if it is true.
Contributions should be refutable claims.
Don’t do: “The rest of this paper is structures as follows. Section 2 introduces the problem, Section 3… Finally Section 8 concludes.” because no one reads this. Use the bulleted list of contributions.
Position: Do not put related work after the introduction. Early related work forms a barrier between your reader and your idea, because most do not understand the discussion in related work (yet). No ‘notational scaffolding’ from the rest of the paper yet, making is harder to discuss related work. Put related work after the details.
It is ok to refer to related work in the paper!
Competition: Do not blatantly downplay. Giving credit to others does not diminish the credit you get from your paper. Be generous to the competition. It is fine to point out on which dimensions your work improves on related work, but also acknowledge weaknesses in your approach.
Main content of the paper, the problem, idea and details, are the easiest to write, because you know a lot about this. But you should consider a few things for the readers.
Presenting the idea: Do not present the general idea first because the readers do not have the same intuitions as you. They will not understand and feel stupid (although motivated readers will keep on reading, trying to understand).
Instead explain the idea as if you were speaking to someone using a whiteboard. Explain the intuition first. Once the reader has the intuition, they can follow the details.
How to convey the intuition? Using EXAMPLES. Only present the general case afterwards. Provide examples that show how your idea solves the problem.
Do not recapitulate your personal journey of discovery, it is not interesting to the reader. Instead, choose the most direct route to the idea. Only point out obvious paths that do not work.
Get your paper read by as many ‘friendly guinea pigs’ as possible, both experts and non-experts.
Each reader can only read your paper for the first time, after that they become ‘immune’ to your paper.
Explain carefully what kind of feedback you want to get. For example, do not care about spelling mistakes for an early draft, but rather where they ‘get lost’, or what was not clear.
One way of dealing with feedback like ‘I did not understand …’ is to explain it on a whiteboard, and from this figure out a better way to explain something. Treat it as an exercise in you learning from them what would be a good way to explain it.
Getting expert help: Send a (final) draft to the competition, asking them “could you help me ensure that I describe your work fairly?”. Often they will respond with helpful critique. If they shoot down your paper, getting this criticism before submitting the paper is better.
But don’t overdo this, they will get fed up and ‘treat you as a parasite’.
Treat every review like gold dust. Be (truly) grateful for criticism as well as praise. Really hard but really important.
Rejections make you ‘bleed’. Your paper has been ‘puked upon by these ignorant reviewers’. Instead of thinking that reviewers are just stupid and missed the point (although occasionally this is the case), it is more constructive to think about how to rewrite something so that ‘even the stupidest reviewer could not have misunderstood it’.
Be grateful to the reviewers because they devote some of their time to read your paper. Thank them in your response and try to take into account what they say.