WARNING: This information is likely out-of-date and will not be updated!
Last modified on 2020-08-19


How to write a paper:





Structure your paper:


Different ways to write a paper:

Try to keep it cose to research standards:


1) Start off with problem space

2) Analyze problem space

- Partition problem space

3) Find correlations between sub-problems

4) Find and analyze dependencies

5) Formulate claim or thesis about problem space


Important step in writing: segment problem state into

key ideas and key problems (just like for research)

Find relations between these key ideas.

The segmentation gives you the necessary structure for

formuating arguments and for structuring the whole paper.


The reader needs to really understand the problem space.

Good papers are those that help you understand a problem

in a different way.



General Stuff:



* Being consistent in your level of abstraction:

Give an abstract view of the problem (big picture)

Subsequenty break down the problem into sub-problems.

Identify key-problems and address them in detail.

What does it take to achieve the big picure?


Paper Granularity:

1) Start with big scope (Grand view)

2) Identify a key problem

3) Explore the problem space


Paragraph Granularity:

Be consistent in the level of granularity.

Particularily important for the design part.


1) Big scope

2) Abstract problem

3) Details


The abstraction detail in each paragraph stays the same.

The next paragraph can introduce a higher level of detail.

Don't mix abstractions in one paragraph.

Think of it as a funnel where you narrow down.


For every sentence and every paragraph judge if it is really

necessary. Does it explain something that has been said before?

Does it help to understand something that is coming? If the answer

to that question is no: leave it.


Don't be redundant. If you need to be redundant you probably

habven't found the right abastraction.


Relate very piece of the paper to other pieces of the paper.

If there is a relation: how does it fit into the paper.




Related Work:


* Don't bash related work too much.

Related work can help you. You show that you know the

community. You don't have to reinvent the weel - leverage

related work to make your point.

Analyze how the related work handles / or doesn't handle the

key problems.






* Section Granularity

Again the funnel (see abstraction granularity).

Try to aid the reader in building the mental model

that you have in mind. Question if everything you think

is obvious is really obvious.

First explain the problem in an abstract way:

1) What is the problem?

2) What are options for solving it.

- Partition the solution space

- Discuss possible solutions

- Pick one and discuss in detail




* Paragraph Granularity:

Give the reader some context. What are we talking about.

Tell the people that you are now describing something:

"We have this thing that does this and that and we cal it Bla".

Hint the reader whether you use a top-down or bottom up desctiption.

Use synonyms and similarities to give the reader context and

to establish some common ground to build on. Add and substract parts

from this common ground. Use clear definitions and possibly math formulas.

Tell the people explicity what the interconnections between components






Opportunity to take the step back. Discuss if your claim is correct or

incorrect. It's not a summary of the paper. Interpret your result.

Make obvious why you found a good solution,


Future work:


"If all you can do with your solution is to port it to windws... why should I care?"


It's not a todo list. Show the big picures. Point out new opportunities.

Take care that it is ambitious but yet realistic. Think about ramifications.

Try go back to the original problem. See if you can generalize the problem.

Generalizing your solution is good future work. Tell people about

current trends in the research are and position yourself in relation

to that trend (we fit into the rend or we contradict the trend).

What will your idea evolve into in 10 years.

- Impressum | Datenschutz -