Writing Papers

Notes on writing papers. Mostly advice I've gotten from Ruofei and Dr. Varshney.

Tasks

Important tasks not to forget

• Check all the references
  • Any symbols in title or accents in names
  • Try to make doi formatting consistent. Either all have URLs or all are DOI:.
  • Try to make sure citations are sorted in \cite{..,..}.

Figures

You can make figures in Powerpoint. However, make sure you change the color scheme from the default so it isn't obvious it was made in Powerpoint and it does not look low-effort.

• Colorbrewer2.org
• Coolors.co

You can also use Google Slides.
Export as PDF rather than a raster PNG/JPEG image. This allows text to be rasterized on the client and images to be full-resolution.

You can use Ghostscript to embed fonts. It's better to embed fonts within each figure rather than in the whole PDF. Reference

gs -q -dNOPAUSE -dBATCH -dPDFSETTINGS=/prepress -sDEVICE=pdfwrite -sOutputFile=output.pdf input.pdf

Video

You should make a 2-3 minute video explaining your project.
This can be done using slides + screen capture or in Adobe Premiere.

Equations

Explain the motivation or derivation of an equation before showing it.

Grammer

Always make a pass on Grammerly.

Common mistakes I make:

• ‘on the other hand’ must follow ‘on one hand’. Otherwise, use ‘however’.
• Don’t use ‘firstly’, ‘secondly’. Use ‘first’, ‘second’.
• Spell out numbers 7 or lower.
• Data is singular and plural. ‘later research’ and ‘recent work’ should use singular verbs.
• List of adjectives need to be followed by a noun.
  • E.g. Existing juice research include: apple-based, orange-based, and pineapple-based juices.

Additional things:

• Always use pural pronouns: e.g. "We did ...", "Our method ...", "Their technique ..."

Latex

• Minor things:
  • Be consistent with spaces in cite lists.
  • Tilde keeps citations on the same line.
  • Avoid orphan lines (last line on another page) if possible.

Paper Layout

Most full papers have the following layout.

Abstract

• Brief introduction of the problem and motivation.
• 1-2 sentences of your method and contribution.
• 1-2 sentences of the findings from your evaluation.

Introduction

This is the most important section for story crafting. The main goal here is to present the problem and convince the reader that the problem is interesting, important, and yet-to-be-solved by prior work or trivial extensions to prior work. A bonus would be to also discuss how your solution to this problem complements existing work or can be applied alongside a variety of other techniques. In your last paragraph, you should clearly list the contributions of your paper and the benefits of your new technique.

Related Works

Some authors put this section near the end but this is just a brief survey of the field.
Make sure to cite all the important papers in the field and some newer ones to show that the area is still relevant.
Try to also cite who you think will be reviewers. Do not criticize other works too specifically since they may be your reviewers.
I.e. Don't say "Li et al. does X but fails to consider Y."

At the end, you can add a paragraph about what advantage of your method has compared to existing works.

Discussion

In the discussion section you can analyse the results from your experiments if you didn't do so in the evaluation section. You can also discuss limitations to your method and potential followup topics. Finally, you can also discuss what additional applications your method might be useful for. However, try to avoid excessive speculation.

Authorship

In applied CS, authorship goes by order of contribution.
Typically the first author(s) is the student who does bulk of the work (experiments, paper writing, presentation).
Meanwhile, the last author(s) is the senior author or advisior who provides ideas, funding, and polishing.
Only first-author papers are used in students' PhD theses.

In theoretical CS (i.e. mathematics), authorship is often alphabetical.

Maximizing Acceptance Chances

• Go through the previous proceedings of the conference you're submitting to and make sure you cite all relevant work. Chances are, those authors will be reviewing your paper and won't appreciate it if your manuscript is missing citations to their work. If there are no relevant work, consider choosing a different conference.