Seminar aus Informatik/Computergraphik - How to write a scientific paper
This document is based on the following resources:
This is a first attempt to summarize the most important aspects of writing a scientific paper in the field of computer graphics. Papers in other fields are usually similar but differ in detail (see links above). If you have any suggestions for improvements, something is unclear or could be formulated in a better way, or have general remarks please report them to the maintainer of this page (see bottom).
When you get your paper(s) from your supervisor and do your literature research you will soon realize that most papers follow a similar form. This should not be seen as a restriction but as help for both the reader and the writer of the paper. The reader is usually better guided to the information he wants (e.g., to get a quick overview of the paper, read the abstract, to know the results, read the results section, simple, see?). The writer, on the other hand, knows how to structure the information so it is best communicated to the reader.
We first state a few general remarks on writing a paper, also scan the above mentioned pages for more information.
- be consistent (also in tense, ...)
- use active form (simpler, easier, stronger)
- never use contractions (e.g., it's, can't, ... )
- all images, figures, tables, ... must be labelled and referenced in the text
- all images, figures, tables, ... must have a caption describing what they are there for
- no paragraph should have only one sentence
- use a spellchecker (however, do not rely on the spellchecker, it does not find every mistake and be careful with "replace all")
- the word data is plural (e.g., data are reported)
- use simple formulations and short sentences, if a sentence can be broken into two, do it.
- try to avoid abbreviations (except: i.e., e.g.), if more convenient, introduce in Introduction
A scientific paper usually is divided into several sections, and each section serves a specific purpose in the paper. Deviations to this structure are possible but should be motivated by a rather strong reason. We now describe the standard format of a scientific paper.
Sections marked with * have to be included. Titles of sections can differ (but should not without good reason), the content however should be there.
The title of the paper is the most often encountered part of any paper and therefore has great importance in the success of the paper. Thousands of readers will scan the title but never read the abstract or paper itself. Abstracting and Indexing services will also utilize the title, therefore, all words in the title should be chosen with great care and their association with other words in the title carefully managed. Titles should never contain abbreviations and jargon. Indexing these word substitutions makes indexing difficult to impossible and impairs the titles credibility.
What is a good title?
The fewest possible words that adequately describe the contents of the paper.
How long should the title be?
The title should be a label and not a sentence. Consequently it does not suffer from the need to be complete and balanced, i.e., subject, verb, object arrangement, etc.
A well prepared abstract should enable the reader to identify the basic content of a document quickly and accurately, to determine its relevance to their interests, and thus to decide whether to read the document in its entirety. The abstract should concisely state the principal objectives and scope of the investigation where these are not obvious from the title. More importantly, it should concisely summarize the results and principal conclusions. Do not include details of the method.
The abstract must be concise, not exceeding 250 words, usually in a single paragraph. If you can convey the essential details of the paper in 100 words, do not use 200. The abstract, together with the title, must be self-contained as it is often published separately from the paper (on web pages, ...). Omit all references to the literature and to tables or figures, and omit obscure abbreviations and acronyms even though they may be defined in main body of the paper. The abstract is only text. Use the active voice when possible, but much of it may require passive constructions. Usually the abstract is written in the present tense. Maximum length should be 200-300 words.
Summarizing, the abstract should contain:
- the question(s) you investigated (or purpose), a motivation for the work (from Introduction) state the purpose very clearly in the first or second sentence.
- A short description of the method (no details), advantages over existing methods
- A brief summary of your interpretations and conclusions (from Discussion)
- lengthy background information
- references to other literature
- elliptical (i.e., ending with ...) or incomplete sentences
- abbreviations or terms that may be confusing to the reader
- any sort of illustration, figure, or table, or references to them
The keywords should not only be taken from the title.
As its name implies, this section presents the background knowledge necessary for the reader to understand why the findings of the paper are an advance on the knowledge in the field. Typically, the Introduction describes first the accepted state of knowledge in a specialized field; then it focuses more specifically on a particular aspect, usually describing a finding or set of findings that led directly to the work described in the paper. In many papers, one or several major conclusions of the paper are presented at the end of this section, so that the reader knows the major answers to the questions just posed.
- state objectives clearly
- why is this work necessary/interesting/useful, motivation
- what are the preliminaries/put your work in context
- cite important literature (if you have no related work section)
This is either included within the introduction or it is a sction on its own. Establish the context by providing a brief and balanced review of the pertinent published literature that is available on the subject. The key is to summarize (for the reader) what we knew about the specific problem before you did your studies (developed your algorithm). This is accomplished with a general review of the primary research literature (with citations) but should not include very specific, lengthy explanations. Objectively state the drawbacks of existing methods and why you have to look for other solutions.
The function of the Results section is to objectively present your key results in an orderly and logical sequence using both illustrative materials (Tables and Figures) and text.Explain used data sets. Figures, Figures, Figures, Tables (timings, space reqs.,...). Compare to other methods. Should clearly describe what was found, and not require the reader to interpret data from figures and tables.
If a figure can be used to show the data, use the figure (e.g. a graph) instead of the table. Most people understand graphs more quickly than tables. Don't present the same data in several ways, choose the one best way. While tables are good for presenting some kinds of data, consider the options. Both tables and figures should be clearly labeled Figure 1 or Table 1 in order of being referred to in the text, and should include a descriptive caption so they "stand alone" without needing reference to the text of the results section.
- reach conclusions about the initial objectives (therefore we call the section "Conclusions")
- show advantages of your method over previously published methods
- state open problems, cases your method does not or insufficiently work
- identify needed next steps in research on the problem
If you received any significant help in thinking up, designing, or carrying out the work, received materials (data sets, source code, ...) from someone who did you a favor by supplying them or funding, you must acknowledge their assistance and the service or material provided.
Do not include your dog, your friends, or wife/husband (in order of importance), or similar things.
- cite all references in the text
- reference order is dependent on publisher, usually it is alphabetically, sometimes in order of appearance in the text
- include all necessary information for others to locate the reference
- usually it is not a good idea to reference web pages
- be consistent
required fields of most common entries (names from bibtex):
- article (an article in a journal):
authors, title, journal, volume, number, pages, year
- inproceedings (a paper in conference proceedings):
authors, title, booktitle, pages, year
- book (yes, a book):
author or editor, title, publisher, year
Sometimes interesting but not necessary information about a paper (e.g., a rather lengthy proof) is included within an appendix. Usually, it is not necessary.