Writing a paper or a thesis in Computer Science is in some ways very different from writing an essay or a blog post. I figured those out by accident and would like to help my readers to be directly aware of them. If you want to get a better introduction, I recommend reading "The Elements of Style".
While writing this article, I noticed the difference between scientific writing and a blog post quite well. Weasel words, for example, make a lot of sense in a blog post.
Most papers I read have the following structure:
Abstract 1. Introduction 2. Related Work 3. The idea ... 4. The idea 5. Experiments 6. Discussion / Conclusion / Future Work References
It should be short. Mabye 700 - 1500 characters.
What is the problem that was investigated? Why is it relevant?
What are other key papers related to yours? This section is a small Literature Review that puts your paper into context.
How did you come up with this? Independent of the experiments, why do you think it is a good idea?
What was the experimental setup? What do readers need to know to make experiments themselves?
Which conclusions do you draw from the experiments? What further work needs to be done?
Types of Publications
- Research Paper: You had a new idea, conducted and reported experiments. This paper presents the original, new idea and the insights you got through experiments about the idea.
- Survey Paper: You provide a detailed overview over a domain. You put the relevant work into context, show how it developed. This is a starting point for new researchers and something that can be cited for "common knowledge".
- Review Paper: You critically analize previously published work.
If you publish the paper to a journal, you could call it an article.
Academic writing is about precision. For this reason, the following words should rarely be used:
- Amount: almost, many, various, very, fairly, several, extremely, exceedingly, few, mostly, largely, huge, tiny,relatively, ((are|is) a number), vast, completely, quite
- Certainty: might, appears to be, theoretically, actual
- Personal judgement: it is easy to see, interestingly, remarkably, surprisingly, excellent, clearly
- significantly, substantially
If you want to give an amount, cite stuff:
- many: Have at least 5 references
- several: Have at least 3 references
- few: Have at least 1 reference
The personal judgement should in many cases be removed without replacement. It depends a bit on the part of the publication. For example, in an introduction or the outlook it might be completely fine and even desirable to have some ideas what implications an observation might have. This also depends very much on the community for which you publish.
When you evaluate a system, you might denote things like the accuracy. Be aware that you also communicate something with the number of digits you denote. The more digits, the more certain you are that this is relevant. So if you say a classifier has an accuracy of 96.123%, then your test data should better have at least 100,000 data points. Otherwise, it does not make any sense at all to denote that many.
Tables and Images
They should be able to stand on their own. Each table and each image needs a small text below / above it, that gives enough context for a reader who is knowledgabe in the area to understand what it says. It is ok to repeat yourself.
Each table and each image should be referenced in the text. Which means that each table and each image needs a number.
- Strunk & White: The Elements of Style (pdf)
- Academic Writing Check: A tool which can detect many flaws in academic writing.
- Difference between Paper and Article for scientific writings
- Literature Review versus Literature Survey. What is the difference?
- Citations with LaTeX