Department of Computer Science
The University of Western Australia

Structuring the Dissertation

The structure entails a number of related aspects, some of which are described in more detail below.

Abstraction Hierarchy

Learning how to convey the essence of your work at different levels of abstraction is probably the single most important skill that you can cultivate for effectively communicating your ideas both in academia and in the workplace.

While the above diagram shows comparative levels of abstraction for the major dissertation components, the same principle should apply where possible at a more detailed level. For example, a typical chapter (or \section in LaTeX) might start with a "leading" or "connecting" paragraph which provides an overview of what is to come in the chapter and how it follows on from the previous chapter. Subsections would then address each of the main issues, with subsubsections (where appropriate) collecting together related details on various aspects of the main issues. This process continues down to paragraphs and even sentences.

Selection of Content

A characteristic example of hierarchical selection of content that should exist in all dissertations is the literature review. It is important to give an overview of the field to show where the approach you have chosen fits in amongst related work. A more detailed discussion of previous approaches should, however, only be provided for those which are directly relevant to your work (such as predecessors or those raising the particular issues you address). Devoting a lot of space to detailing less relevant approaches detracts from the space available for presenting and discussing your own work. This may result in "selling yourself short"!

In Computer Science terms selecting material for a dissertation is a fixed-resource (35 pages) optimisation problem, and you want to maximise the fitness of your dissertation. Using abstraction effectively means that you can have the best of both worlds - discussing the bigger issues and devoting adequate space to your work.

Providing a Framework

"Tell them what you're going to tell them, then tell them, then tell them what you told them."

so many times that you simply filter it out as your supervisor waxing lyrical again. The importance of this rule cannot, however, be overstated.

While you (and possibly to a lesser degree your supervisor) have a complete picture of your work in your head, it is important to remember that other readers are starting from scratch. You must therefore provide them with a framework that they can fill in as they go through your work. For this reason the first four components in the diagram above - the title, abstract, introduction and conclusion - should all be abstractions of your complete story.

The introduction should also give a more concrete roadmap of the dissertation, saying what can be expected in each chapter. This will typically be provided in the last subsection of the introduction with a title such as "Structure of the Dissertation".

It is easy to underestimate the importance of your title and abstract (and even introduction) and just "get something down" as they are not your "real work". In some senses they are the most important part of your work because, apart from providing those who are definitely going to read your work (such as markers) with the structure (and an initial impression), they provide the basis on which other readers decide whether to read your work at all.

Telling a Story

The literal meaning of a thesis, as given by the Hypertext Webster, is :

1. A position or proposition which a person advances and offers to maintain, or which is actually maintained by argument.

You should state your position or proposition in the abstract and introduction, provide the details of your argument in the subsequent chapters, and tie it all together in your conclusion!

Terminology

• Technical terms relating to the area of study, and terms that are not in common usage or are used differently from their common usage, should be defined or explained at their first use. (This does not necessarily apply to the title and abstract.) It is conventional to write terms being introduced and defined in italics.

Paragraphs

• Do not begin paragraphs with joining words like however and but.

• Do not begin paragraphs with sentences that rely on the context of previous paragraphs to make sense. For example, rather than starting a paragraph with
  In order to achieve convergence it was initialised to...

  start with

  In order to achieve convergence the learning algorithm was initialised to....

Sentences

• Beware of tautologous phrases. These can be simple, such as
  ...we reran the algorithm again...

  or more subtle such as

  ...as the problem becomes larger the space of solutions grows at a rate that is exponential in problem size....

• Do not split adverbs and verbs around a noun phrase. For example
  Roughly the result from the search approximates a...

  places a higher cognitive load on the reader than

  The result from the search roughly approximates a...

  since in the first case the reader must push "Roughly" onto his or her cognitive stack while processing the noun phrase, and then pop it off when "approximates" is reached.

• Avoid repetition of key words (that is words other than "the", "that", etc) and phrases in sentences. For example, rather than
  We look at the problems involved in looking at search as a...

  write

  We discuss the problems involved in looking at search as a....

• Avoid using three or four words where one will do. For example rather than
  ...the testing of the algorithm could be conducted by...

  simply write

  ...the algorithm could be tested by....

  As well as saving space, this makes the text much easier to read.

General Grammar

Tense

Beware of switching between present and past tenses. It is usually safest to stick to present tense.

Plurality

Beware of changes in plurality in sentences and paragraphs. For example:
The main problem that can be encountered when using optimisation algorithms is problems with the inaccuracy of the fitness function.

mum and ma

In words such as "optimum/optima" and "datum/data" the mum form is singular and the ma form is plural. For example a function may have a global optimum and many local optima. Also in this case "optimal" is the adjective - thus we would have an optimal solution (as opposed to an optimum solution).

Non-Discriminatory Language

The use of non-discriminatory language, including non-gender-biased language, is University Policy (see Section 6, Part 4 of the Human Resources Policy and Procedures Manual). The University provides a guide on the use of non-discriminatory language to assist staff and students.

References

• Where URLs are included for Web references you should include the most recent date on which you have accessed the page and hence verified its availability.