How to Write a Great Research Paper

I am going to speak about the act of writing,

which is something every
researcher has to do a lot

and the good news is
that some pretty simple things

can make your
written output a lot better.

So that's why I titled this:
Seven simple suggestions.

So here we go.
Seven simple suggestions.

Here's the first one:
Don't wait: WRITE.

This is what mistake,

what a lot of people make.

Here is a typical plan for doing research

Number one:

Have a brilliant idea.

Number two:

Spend months doing good research to back it up

Number three:

Write the paper.

In last two weeks before the deadline for the conference.

This is a bad plan, right.

Here is a good plan for doing research:

First: have an idea.

Second: start writing the paper.

Third: use the paper as a forcing function

to make you do the research that articulates the paper.

Ok?

See how diferent it is?

Right?

It means you start writing very early.

Why is this good? Well.

I often find, if I spend invested months of effort

hacking around and writing a code and thinking

and I start to write the paper,

than I relise that some of my previous work

was misdirected, right?

It wasn't able to do useful goal,

certainly not for this paper

and I also discover

some key parts of the paper that need more work,

right.

So you wanna use the paper as a forcing function to learn that early.

It also gives you a good way to communicate with another people

Research is all about communication. Right?

If you...Maybe you are kind of person who likes work just on their own

in a windowless room with no lights.

But I like to work with other people

I spent a lot of time at the white board

If you have something written you have a new mechanism for communicating with them

Right?

Is very important is a mechanism for communicating

That's all along way saying

that really writing is not the way in which we just report research

for me is a way which I do research

and I think you should think about writing in that way.

It's not just the output, it's a computation

It's the stuff that makes research happen

And it's because somehow we think more clearly when we write, than when we just think.

At least I do.

So.

Ehm. One problem with this plan here is that it starts with this idea bit.

Right?

Where this idea come from?

And so..ehm...it's tempting when you looking at another peoples' work

to think "oh my" everybody else has very clever ideas and I am a mere worm

and I have such a trivial boring ideas that nobody would be interested in them

And do you know that feeling? Yeh?

You sit there and you don't feel very creative

Let me tell you that this is what every researcher feels most of the time

There are days when you have miraculous breakthrough

and you really know your God is in heaven and you know that all is right with the World.

But most of days you thing I am worm this is the natural state of the researcher.

Right?

So what you have to release is that even you know [Tarjan], and so forth, thinks that he's a worm most of the time

and what what the researcher do? Good researchers

is they simply start writing anyway, right?

They write a paper about any idea no matter how trivial or insignificant it may be.

My experiences that the clever the research student then more they prone to this failure mode

either they don't understand something which case they are depressed

well stay do undestand something which they do believe, in which case they think it is trivial and nobody would want to know about it, all right?

So in both cases depressed.

This is not a good situation, right?

So just write an idea no matter how insignificant it seems to be

because my experience consistently is that when you write the paper

your idea develops and ramifies computer science is like a snowflake or flower

you start with a little seed and its ramifies ahead of you in an interesting things

something looked boring turns out to be actually rather interesting

not always sometimes you start writing the paper and it turns out indeed to be weedy and insignificant

and maybe when you published it and it will be done quickly and you put it on your homepage, all right?

So, write early!

And, and and I really do believe that the second thing is much the most a common case.

It almost invariably turns out to be more interesting than you thought.

All right, number two if you gonna have this idea and wrote about it you need to be clear what it is.

The business of writing a paper is to convey from your brain into the minds of your readers your idea.

So think of your papers like, like a virus, right?

You trying to, you try to infect your readers' minds with your idea.

And then you will infect them and then they talk to another people and they will infect them

So it's like a kind of contagious play which is going to sweep the world

And everybody is been thinking about your idea because it is so well infectious.

Right?

So...ehhmm....my, my, my knowledge here with, with Mozart

It's just houndred of years after the Mozart died we are listening the people read his papers.

Or more precisely play his music.

Where you go to concert hall to hear his papers because

their ideas were so infectious

Don't you think that's amazing?

Won't it be remarkable that in four hundred years time people will still reading you papers.

That's not all likely

But I think that's the..,that's a kind of idea like to get

your your..the papers is not mechanism for getting promotion

It's mechanism for convey ideas from your head to somebody elses head

If you don't convey them, right, if don't bother to tell anybody about your ideas,

then your minds don't have them

Right?

Even if Einstein had sat in the window this box not tell anybody about relativity.

Then we wouldn't know relativity, right?

That means you need to know what your idea is. Right? So when you write a paper in the end at least by the time you finished.

You must know what is the idea that your paper conveys.

It's surprising how hard that could be to determine from people's papers

As a viewer if you read somebody's paper when you finish you can say:

"What idea did that paper convey?"

If you find that hard to articulate as a reader, then you know, in your view, you can say I really couldn't figure out what this what the idea this paper was about.

And so than you apply to your own papers, right?

It's actually sometimes quite hard to know exactly what your idea is to begin with.

But you must now by the time you finish.

Eeehrrr...if you find that yourself thinking, ohh..I have actually three ideas in this paper.

Than what you do? Just write three papers.

Right?

It's wrong to try to merge them all in. In which case each one becomes cryptic and incomprehensible. Which you try to squidge into ten pages.

Just write three papers, that's cool, right?

That's not salami slicing, that is taking ideas and expressing them.

A good idea in your paper is to say explicitly when you get to main idea

It's suprising how seldom is this happens.

So I often try to write phrases in my paper that say the beginning of section three somewhere the main idea of this paper is this..

Because I have to explain some context and set up and background.

And I wanted to be absolutly clear when I move from describing the context saying: here is the payload.

Right?

The viruses is about to arriving in your brain. Prepare!

Does it make sence? All right?

If you don't say that, if you leave you reader to be a detective.

So, subsequently they have to reverse engineer what they think you meant was the key idea. That's not good, right?

Why not said it explicitly, be completely up front about this.

Third thing: Tell a story!

So if you know write a paper this applies I am focusing mainly on papers.

But I do also mean dissertations and so forth. Everything applies everything you like really.

Is tell some kind of a story

I always try to imagine when I'm writing a paper, that I am standing a front of the white board and explain it to a colleague.

It's amazing how differently people present things at a white board and they will in a paper.

Right?

Without white board they start away ..??..with examples and explaining quite diffrently than if they..if they will not be out there..you you you get the idea..

They often explain in much more accessible and engaging way at the white board than in the paper.

So if you want to by accessible and engaging ......... do the same in the papers you do at the white board. Nearly.

You need a little have a bit more substance.

So here is a narrative flow that I usually try to follow with my papres. Right?

You want to say here what a problem is, you want to motive your readers to say way it is an interesting problem

You want to say at least briefly why it is unsolved problem, so be aware of solving, it is unsolved.

And then you want to present your idea, that's the payload, right now

Right?

And then you give quite a bit of detail

about exactly how what your idea is and how it works

Ehm... Then you want to say something about

how your idea compares with other people.

See, you are kind of trying to lead people in

You know, no everybody will read

right the way through your paper.

Your ideal is

That whereever anybody stops reading

they take away something valuable with them.

And moreover every bit they read

makes them want to read more, right?

This is back to make it accesible, right?

So, you know they say:

Ah, here is an interesting problem

I wish I could solve that one!

I wonder if he can solve it!

So, you know - they got a hook!

They incline to read some more.

Ehm, here is my - so, the typical outline

for most papers that I write

and indeed it works for dissertations as well.

A short abstract.

An introduction.

Then something about the...

stating what the problem is.

Then a quite a bit of

a short piece explaining what the idea is

kind of intuitively

and then a longer piece explaining the details

behind the idea

that sort of fills out the evidence.

Then something about relaive work.

And I am going to say a bit more about

each of these sections

in the following piece

but this is my picture of

how the structure of a paper might go.

And look at the numbers of readers, right?

More people read your abstract and title

that will read all the rest of the paper

so each time you want to give them a hook

to continue.

Here is what I think

your introduction should do.

One page!

Describe the problem briefly

and articulate what your contributions are.

So, here is some by way of example

Ehm, describing the problem

I would suggest you introduce your problem

with an example.

So, my good way of doing to say:

is there any typewriter fonts on the first page?

I am a programming languages guy

and programming languages people tend to

put example programs in a typewriter font.

So, I know if there is a typewriter font on the paper

it's probably an example.

That's good, right?

But in your fields it may differ

but anyway, start with an example

that'll ilustrate your problem.

So, here is an example of

a paper that I wrote some time ago

and the very first thing I did was to give

a little program and explain a problem with it

and explain that somehow

my paper is gonna fix this problem.

OK?

So, that's an initial hook.

Not a general description of a problem