Scientific writing - ftsrg/cheat-sheets GitHub Wiki

Reading

Implementation

Writing

How to write a good introduction:

  • The Introduction provides the first impression about the paper. A good structure recommended by Stanford InfoLab's Jennifer Widom in her article about Tips for Writing Technical Papers:

    • What is the problem?
    • Why is it interesting and important?
    • Why is it hard? (E.g., why do naive approaches fail?)
    • Why hasn't it been solved before? (Or, what's wrong with previous proposed solutions? How does mine differ?)
    • What are the key components of my approach and results? Also include any specific limitations.
    • Summary of contributions in bullet form, doubling as outline for the paper

  • Bugs in writing

How to break a title

Hungarian documents

Punctuation

Punctuation Guide

Journal papers

Writing journal papers is a long process - it often requires months to prepare the experiments, draw the figures and write the paper. We gathered some tips to make the process a shade easier.

Approach on writing a journal paper

  • If it's possible, start with a conference paper, gather some feedback and incorporate them to the paper. Many conferences invite some submissions to journals, which usually has multiple benefits: 1) you have to fiddle a bit less on stating your contributions w.r.t. the conference paper ("you need a 30% delta") 2) the review process is usually a bit faster 3) you get a hard deadline, which helps in scheduling your work. However, some papers do not work well as conference papers, as the Dealing with rejection post says:

    Some papers are simply not fit for the format of a conference, and may benefit from being published at a journal. This is certainly true for many theoretical model papers or literature surveys, which do not necessarily need (or lend themselves to) a public talk at a scientific event. Some papers are too complex or too long to be conveniently discussed at a conference, and instead require the back-and-forth of the journal review cycle, as well as the additional page allowances of a journal.

  • The review process for journal papers often include multiple iterations with the reviewers (major revision, minor revision, etc.). Hence, in many ways, your first journal paper submission is more of a "beta version" than what you submit to a conference. This means, that if you are short on time (or just want to get the process started), you do not have to be meticulous on all your figures and plots. During the review round, you will have the chance to improve the figures and you will probably have to re-run the performance experiments with the changes requested by the reviewers.

  • In addition to the previous point, also remember what Knuth et al. state in their excellent report "Mathematical writing":

    Journal articles should be polished and timeless. Conference papers can be a little rougher. Conference papers are appropriate for work that is “not yet ready for the archives.” Technical reports (usually distributed by an institution) are good for work that is not even ready for the general world but still should be written up.

Technical details

  • Check if colour prints in the journal cost any extra. If they do and you do not want to pay the extra fee, prepare the paper so that the text is black and white only. For example, do not use the textcolor command of the xcolor package. (I am not sure about listings, but you should prepare them so that they are readable in b/w print). Even if you do not pay colour prints, your figures and plots can use colours but you should check if they are readable in b/w.
  • Some publishers (e.g. Springer) generate a "PDF proof" version (with watermarks, line numbers, etc.). This conversion does not always go smoothly. To avoid any surprises, upload an incomplete version well before the deadline and check the output. (See the related section in Publication checklist).
  • Check if the journal provides copy-editing. If it does, do not fiddle too much with tables - they will be reformatted anyways. Also, either read the style guide or if you're too lazy, at least check some accepted papers for style hints. For example, do not use headline capitalization for Springer journals. If your paper is accepted, it will be copy-edited and copy-editors will change to the style prescribed by the journal. So deviating from the style guide will only make the diff noisy during the publishing process. Typical style-related decisions:
    • headline capitalization
    • abbreviating Section/Sect., Figure/Fig.
    • using a full stop at the end of paragraph names (\paragraph{Something} vs. \paragraph{Something.})

Citations

Use ISO-4 abbreviations for journal titles: