How to Write a (Good) Paper

#PGBBS - Konstantin - 10.11.16

  • “Regeln sind da, um gebrochen zu werden”
    • Nicht auf biegen und brechen beim Paper schreiben.
    • Es gibt einige Regeln, teilweise auch community-driven, die fürs Schreiben eingehalten werden sollten. Einige dieser Punkte werden hier aufgeführt und diskutiert.
  • Top to bottom Ansatz: Science - Paper - Sections - Paragraphs - … - Sentence
    • Viele der Regeln lassen sich anwenden, und zwar sukzessive von oben nach unten, siehe z.B. Bonbon-Regel
  • Don’t be to fancy, we are no poets
    • Small vocab, few expressions etc
    • To repeat words is ok, if the meaning needs to be there
  • BUT: Be fancy in your scientific ideas!
    • Stick to the rules
    • Stick to the page limit
    • Challenge is to get your info into the defined borders (pages, or time), which is enough. If you need way more space, than your explanations might not be concise enough
    • Example: Haikus (japanese poetry)
  • Constraints and rescrictions fuel creativity
    • Might be hard to use sometimes
    • But you should try to bring the elements into the restricted form - the results might surprise you!
  • Contributions!
    • First impression for the reviewer is important
      • Sidenote: The reviewer in general only reads the abstract, title, intro, and conclusion
      • You have to catch him with these things!
    • Make your contributions clear, also visually. Some reviewers want to see the contributions right away. You can use for example bullet points with your contributions to make them stand out of the ordinary text.
    • Catch his interest to read further
  • Mathematicians and Computer scientists cannot write - Hey!
    • Be careful to not fall into a very schematic description
    • e.g. mathematicias: sentence, sentence, rule, definition, and after these four steps you finally get the (mostly only important) result and thing to read
    • Problem driven writing
    • Start with the problem, get to the solution
  • Storytelling 101 - Bonbon Modell
    • IN - BODY - OUT
    • “Tell them what you are going to tell them, then tell them, then tell then what you told them”
    • IN: problem setting solution, pick up the reader at his current knowledge, do not assume previous knowledge
    • BODY: Tell the process of how you solved the problem and what was used for it
    • OUT: Again refer to the problem and that your solution solved the problem
    • Example story: Classic western - build tension -> epic shootout -> clear the scene with nice conclusion
  • Write the bonbon recurseively
    • Science - Paper - Sections - Paragraphs - … - Sentence
  • Budget your text, every section and sentence has its own length requirement
    • Example of the 12-pager: 3 pages + 5 pages + 2 pages + 2 pages
    • Related Work Budget: Do not describe the work others have done in more pages than your own work!
  • Two mantras
    • “The point, the point, get to the fucking point!”
    • “The crap, the crap, cut the fucking crap!”
  • Adjectives are bad!
    • Mostly only subjectives and verbs are important
  • Position of related work, can be at front or in the back
    • Manu: the position might correlate with the question if you want to position your work against the others, to highlight things you have done better
  • Title
    • keep it short, but cover the main points and the contributions!
    • avoid “towards…” :’(, “on…”

… skipped some things because of time reasons :) …

  • Sentences
    • Avoid passive speech, actives are way stronger
  • Paragraph lead overs
    • Dont link to the previous paragraph, but use the last sentence of the preceding one to lead over
    • This makes every paragraph start way smoother and it is better to read overall
  • Use Present Tense (Exception Conclusion)

  • No Cascading subsentences (Schachtelsätze)
    • There are people who never use “which” => Which-hunting.
    • Other nono-words: very, a lot …
    • Remove insertions (Einschübe)
  • Never copy - but imitate!

  • Grind and polish. Read pharagraphs out aloud.

  • Pictures good! Formulars bad!
    • Picture explanations are important
    • Manu: Oftentimes I have had bad comments about my pictures that are not described enough
    • Manu: Sometimes you can use caption to explain the picture

APPENDIX yet another skeleton for a computer science paper

  1. Introduction
  2. Background
  3. Approach
  4. (Key) Implementation (decisions)
  5. Experiments
  6. Related Work
  7. Conclusions


Quotation from Medawar1979, p. 66:

A summary must be complete in its own limits. It may well start with a statement of the hypothesis under investigation and end with its evalution. Nothing is more abjectly feeble than to write some such sentence as “The relevance of these findings to the etiology of Bright’s disease is discussed.” If it has been discussed, the discussion should be summarized, too. If not, say nothing.

He makes two points:

  • abstracts should be self-contained
  • generally, avoid empty phrases