INFT 4017 CIS Research Methods

Getting started with writing your thesis

In this lecture and workshop, we will focus on getting started with writing your proposal and thesis. We will focus mainly on the literature survey, although will look at the entire structure of your proposal and the subsequent thesis.

The structure of the whole thesis

The thesis must "tell a story" with everything included obviously relevant and nothing irrelevant or unnecessary (like overdoing the background, e.g. talking too much about supporting technology).

There is a fairly standard structure for a thesis, that you will see often in published papers from your bibliography as well. This structure generally looks like this:

As discussed last week, there is a diffrence between the background and literature survey, although sometimes they two can be hard to separate. Just think of the background work as that which solves a different problem while related work is trying to solve the same problem.

Also we mentioned in the last lecture that is is worth considering whether your literature survey could form the basis for a paper to be submitted to a survey journal such as ACM Computing Surveys.

The Research Proposal

Your Research Proposal is the first stage of your thesis, and its aim is to describe the work you are going to undertake in your project, justifying it firstly in terms of motivation (why bother doing this work at all) and setting the context (the background and related work).

It is usually submitted at a relatively early stage of your research, in UniSA at around the 6-month mark (for PhD and UG/PGT projects). One of the purposes of the research proposal is to create a document of the work of your research, so that others can assess it. This assessment is to ensure that your research proposes aims that are achievable in the timeline, and that the deliverables will be of the standard and type required for your thesis.

The Research Proposal also has the benefit of making sure you have a certain amount of work completed by the 6-month mark!

For the research proposal, you are expected to have a formal question or questions you intend to answer with your work, and to have proposed a set of activities and/or a method for discovering the answer to your question.

In some respects your research proposal is different to your thesis, as you have not actually done any of the work you have planned. At the proposal stage, you are only expected to have justified the planned work and describe the plan for achieving your aims.

However, in terms of the justification, this is essentially the first part of your thesis, and you will probably not significantly alter the motivation or literature survey between the proposal and the final thesis. Thus you get to reuse quite a lot of the research proposal in the final thesis. Probably the only material from your proposal that will not also be incorporated into the final thesis is your schedule.

So, in your proposal, you would expect to put the following sections:

As you can see, your proposal will essentially be the first part of your final thesis.

A good example of a research proposal is the PhD proposal submitted in October 2008 by Gavin Smith.

The Literature Survey

Now let's look at the literature survey.

This is a critical part of your research proposal and your thesis as it demonstrates your knowledge of the similar research in your area, and puts your work in context. In a way, its entire purpose is to set the stage for your own research, by pointing out what has already been achieved, but also by pointing out what has not yet been achieved. If you write your literature survey well, you can lead the reader to draw the conclusion that your work is necessary even before you announce it yourself.

Additionally, you can sometimes present your literature survey in a form that is itself a novel piece of work, perhaps by considering your field from a novel point of view, or creating a framework for understanding and analysing other work.

Frameworks or formal models can sometimes be an excellent way of analysing a collection of related work since it offers a rigorous method for comparing them - you can say how each work fits into the model (for example what values it has for certain variables) and then you can easily compare the differences between different related works, and most helpfully you can also then see what the other related works may have in common with each other or with your proposed work.

Whatever you decide on, you need to ask the right questions about the related work, taking a point of view that is specifically tailored to the project you are taking. Let's look at a recent example of the kind of questions to ask:

The project being to cluster objects that are "linked" according to co-selections, rather than having similar content. Modelling the data as a doubly-weighted graph.

Collect information on all clustering algorithms, initially consider them from the point of view of
i) what is being clustered (documents, text, images, spatial whatever)
ii) what feature is being used to cluster them (content, links, clickthroughs)
iii) what is the purpose of the clustering (grouping semantically similar documents)

Later try to cast into a single framework for comparison 
What we want to do is identify which of these is suitable for clustering

Clickthrough-based clustering is essentially *implicit* content-based clustering - this may be a 
direction to take later i.e. to measure the value of implicit content analysis by humans versus algorithmic content-analysis.

As you can see from the above example, the method of comparison can give you a method of evaluating the value of your work. For example in the above example, it would be sensible to evaluate the value of the proposed clustering algorithm/method by comparing its performance on exactly the same set of objects with a content-based algorithm and a link-based algorithm.

Note that the evaluation of your work is every bit as important as the work itself, because it proves that your work has value. It is not quite the same as the literature survey becuase the literature survey points out ideas that have not yet been tried out, whereas the evaluation is where the testing of ideas takes places.

Let's get started on it

For the rest of the time, we will start on our proposals, top down!

Take the structure of the research proposal, and start creating your own sub-headings and maybe even sub-sub-headings. Write some notes about what you plan to write about under each sub-heading.

Focusing on the literature survey, see if you can think of some characteristics of the related work you have read about (from your annotated bibliography) and try to group or segregate that related work according to these characteristics (do this by listing the publications in the appropriate group).

If you can think of more than one characteristic for grouping and segregating related work, that is good.


Last update hla 2009-04-08