One important part of academia is to write papers about your work. Looking at the paper output of a academics is a common way to judge their “value” as a researcher. There’s even an expression to that effect: “publish or perish”.

But how do you write a good paper?

Let’s have a look at how Professor Simon Peyton Jones from Microsoft Research sees it:

He makes seven points, and they are all very sound. You should really watch the whole video, but I have tried to summarize it below.

Let’s have a quick glance before I go into more detail:
1) Write your paper early
2) Identify your key idea
3) Tell a story
4) Contributions: early
5) Related work: later
6) Put your readers first
7) Listen to your readers

Title, Abstract (4 Sentences), Introduction (1 page), The problem (1 page), My idea (2 pages), The details (5 pages), Related work (1-2 pages), Conclusions and further work (0.5 pages)
Professor Peyton Jones’ preferred paper structure (including page count and expected amount of readers)
1) Write your paper early
One common approach is to first have an idea, then to do the research, and then write the paper. There are two obvious problems:
a) You often don’t have a lot of time for writing the paper after the research (because the next research activity already needs your attention, and because the conference deadline is near).
b) When you actually write the paper you notice that you should have included this or that in the research or should have done things differently.
If, however, you start writing the paper immediately, then it accompanies and informs your research from the beginning.
He also argues that you do not need to have a “brilliant” idea in the beginning. Just start with a small idea. It will grow and develop over the course of the research (and the writing).

2) Identify your key idea
“You want to infect the mind of your reader with your idea, like a virus.”
In order to do so, a paper should have just one idea (if you have more ideas, just write more papers). Therefore the paper should be focused, and a reader should always be clear about your idea throughout the paper. It also helps if the paper includes phrases like “The main idea of this paper is…” or “In this section we present the main contributions of the paper.”

3) Tell a story
Professor Peyton Jones compares a typical paper to presenting your idea on a whiteboard – which usually people do in a very different way. The whiteboard narrative goes something like this – and should be followed in a paper as well.
a) There is a problem.
b) The problem is interesting because…
c) The problem has not been solved yet.
d) This is my idea.
e) The idea works (details, data)
f) Comparison of my idea with other people’s solutions/approaches

4) Contributions: Early
Do not hide the contributions at the end of your paper. Instead, they should make up an important part of your introduction. Write them already in the beginning (and be open to editing them later). Bullet-points work well for this, and you can also use them to reference later sections of the paper (instead of writing a boring paragraph like: “In the next section of the paper we will…”)

5) Related work: later
Professor Peyton Jones criticises that many papers have a related work section right after the introduction. His opinion, however, is that the reader is not really interested in the related work at that point. They are either expert readers, who know most of the related work already, or they are no specialists and then might have problems understanding this section. Also, because your idea has not yet been fully introduced, it is very difficult to compare the related work with the approach presented in the paper. He therefore suggests to move the related work section to the end of the paper (see structure above). This allows you to closely connect your own work with that of others.
He also gives really good advice how you should talk about specific works: praise them where praise is due and acknowledge weaknesses in your own research.

6) Put your readers first
His main advice here is to give the “intuition” about your idea first. Do this by providing an example and only then describing the general case. Also, you should not mention the dead ends that you encountered on the way. This might be interesting for you (“soaked in your blood”), but it isn’t really for the reader.

7) Listen to your readers
It might sound obvious, but in order to find out what the reader thinks, it is good to actually have readers. Before submission. Remember that you can only “use” any reader once (as afterwards they already know what the paper is about). It is also more important to find out where they got “lost” and not where you have spelling mistakes. So make sure to tell them that you want to know where it was difficult for the reader. And then afterwards, try to explain it to them directly – and if that explanation works better, but it in the paper.
Finally, even if reviews are harsh, try to understand what caused the problems.

And that’s it!
As said, I think his talk includes excellent advice. As I am about to start writing a paper now, I will try to take his ideas into account and see how it goes.

How to write a great research paper
Tagged on:

Leave a Reply

Your email address will not be published. Required fields are marked *