Thesis writing: context or literature review?

Question

I am currently attempting to write up my doctoral thesis, which covers a couple of domains. It's heavily involved in farming data, but the PhD itself is in computer science. Naturally I start the thesis with the introduction, which talks about the problem definition etc. However, the following step is proving problematic.

It seems typical for a literature review to come after the introduction, but a literature review of... what? In order to ground the research the context must be discussed, and the research surrounding that context. This seems particularly important when the examiners are unlikely to be familiar with the farming context that the research is based. However, this leaves me with a structural problem. It seems odd to go from talking about terms relating to dairy farming, and research about cattle, to talking about particular algorithms.

Moving from one domain to an entirely different one within the same chapter seems conceptually jarring. "... and that concludes the discussion on milk yields. Machine learning has a lot of uses in industrial fields and the focus of much current research...."

I haven't heard of a separate chapter for research context and literature reviews on techniques though... is this something which is "done"?

Answer 1

I was bothered by that as well when I wrote both my MSc thesis and my PhD dissertation. I get the impression that it is considered acceptable to have a selection of unrelated topics in the literature review without any context to tie them together. The reader will see the context for a topic when it is referenced later in your thesis. At least, many of the theses and dissertations I read were structured like that.

However, I myself was not happy with this approach. The problem was especially bothersome to me, because my dissertation pulled together ideas from a broad range of disciplines. So here are some of the techniques I used to "set the stage". These are just ideas; you don't need to do all of these.

• In the introduction, include a few sentences mentioning that you will be drawing on ideas from fields X, Y, and Z.

• Include a topic map that illustrates how the ideas relate to each other. Below is an example from my MSc thesis. My advisors were a little wary of this "weird diagram" at first, but they warmed to it. Later they asked for a similar diagram in my PhD thesis.

• Try to arrange the topics in your literature review in some sort of logical order, and add sentences that provide a segue from one topic to the next. For example, at the end of a section on cognition and before a section on evolution, I wrote "The next section presents a framework for analysing cognition within the context of evolution."

Answer 2

The only universal rule with your thesis is that you, your committee and your advisor must be happy with it. You can do whatever you want so long as those requirements are satisfied. The corollary to this is that there's little point asking such things here, because it really depends. You should ask your advisor and/or committee.

By way of general advice, the introduction of the thesis should convey the goal of the research and the basic information needed to understand what is being done. Unlike a paper, the readers may be very unfamiliar with your area of work; you should therefore establish the basics from the beginning. You don't need to write an encyclopedia - citing relevant reference material and giving a high level summary is usually sufficient. But again, it depends on your committee.

In your case, you can probably expect that any reader knows what a cow is. However, they may not be familiar with how agriculture actually works at scales you are analyzing. At least a few paragraphs describing what sort of farming you mean is a good idea. You also want to clearly explain the processes and quantities that are represented by the data you analyze. Lastly, there should be an explanation of the relevance and usefulness of your analysis to the farming sector, if any - but that can go into the conclusion.

Conversely, a reader knowledgeable in farming but not computational analysis will need some grounding in relevant literature context.

Presumably you have some papers you published on this subject. How did you write the introduction for those? That's usually a good starting point. If you have written no papers, there must surely be at least papers you've read. And of course, you are probably not the first person to have written a thesis about using machine learning in agriculture.

"... and that concludes the discussion on milk yields. Machine learning has a lot of uses in industrial fields and the focus of much current research...."

You don't need to say it concludes it. The final sentence, by virtue of being the final sentence, obviously concludes its section. Then you make a section break and title it "applications of machine learning in industrial fields". Texts can make more than one point, it doesn't have to be a single unbroken chain of reasoning.

I haven't heard of a separate chapter for research context and literature reviews on techniques though... is this something which is "done"?

It's quite common. Sometimes people have already written review articles and they even recycle that into the thesis. If length is a concern, you can have a whole chapter, or even multiple chapters, devoted to the literature review. But often it is unnecessary - I'd check with my committee.

Answer 3

First of all, as with all questions regarding theses, what ultimately matters is what your supervisor and doctoral committee is comfortable with. So, my answer (and any other answer that you might find helpful) must be understood with the caveat that it is subject to what is acceptable to your thesis committee.

With that understood, you are asking about a literature review. So, the primary point of clarification to help you decide what is best is: which literature do you hope to contribute to in your thesis? That is, is your thesis only a contribution to the computer science literature, or do you also hope to contribute to the dairy farming literature? Note that I am not asking about contributions to dairy farming practice (let's take it for granted that your thesis is a practical contribution to dairy farming); I am only focusing on the question of contribution to dairy farming scholarly literature. From the details in your question, I think we can take for granted that you want to contribute to the computer science literature, so let's focus on the dairy farming literature question.

If you hope that your thesis will also be a valuable contribution to scholars of dairy farming, then you should certainly review the relevant literature so that such scholars can appreciate where you contribution fits in. I would think that, in this case, you would focus on two main kinds of studies:

• Algorithmic approaches to solving problems in dairy farming: here you would remain in the scope of dairy farming literature but focus on any uses of computer algorithms to solve any problem in the domain.
• Similar algorithmic approaches to yours in animal husbandry: here you would broaden your scope to any topic in animal husbandry, but you would limit the algorithms that you review to those that are similar to what you will present in your thesis. (I don't know what "similar" means: you are the one who knows what that might mean because you know your algorithms best.)

Whichever of these approaches you adopt, then you could have two major sections in your literature review chapter, one for each literature domain (that is, one for computer science, the other for dairy farming). In addition, your abstract, introduction chapter, and conclusion chapter should devote ample space to presenting and discussing the contributions to dairy farming literature, in addition to the computer science literature.

If, in contrast, you are only interested in contributing to the computer science literature and your contribution to dairy farming is mainly practical, not mainly scholarly, then you do not need to do any sort of literature review of the dairy farming literature. All you need is a background explanation of the practical context so that computer scientists who read your work can understand the practical domain of application of the algorithm. For this, you do not need to cite any scholarlly literature at all. Scholarly literature is mainly to document new, proposed cutting-edge suggestions--it is not primarily for describing common practice.

In this case, it is perfectly acceptable to write one page or more of necessary background information off the top of your head, including only the dairy farming information that is necessary for a computer scientist to follow the rest of your thesis, nothing more. Then you should probably add one or two references to more detailed sources for computer scientists who want to know more about dairy farming, but these references should probably be guides for the general public, not scholarly literature, since the goal is not to describe the state of the art but rather to give the most useful background. For example, if it is sufficient, a reference to the Wikipedia article on the topic might do the job.

In this case, such background pages would only take one to three pages in the introduction chapter of the thesis. There should be nothing at all in the literature review chapter. Then you should probably add a few paragraphs in the conclusion chapter talking about the practical implications of your algorithm.

Answer 4

Write two or more different introductions, long or short, depending on what you want to emphasize. This makes it easier to write and adds explicit structure, which generally makes it easier to find things. Your thesis will hopefully be read more for the information than the narrative.

These can be sections, subsections, chapters, or whatever feels natural. You can start a chapter by explaining that the literature review or introduction is organized in a particular way. If you feel uncertain or if it is non-standard in your field, then you can explain why you are doing it. But I would suggest just doing it; people will see that it makes sense and be happy.

Background for applications in farming

We are interested in the following questions: blah, bløh.

For an introduction to blah, see textbook/review article. For our purposes it is sufficient to know this and that.

The issue with bløh is something. For further information, consult the books/review articles.

Prior computational studies

We mostly use method C. It is fairly new and has been studied in the contexts of D and E. Method C works by...

Computational studies in farming

Completely different computational method M has also been used in studies of farming, but for quite different purposes.

Answer 5

I think it is fine. Kind of cool to have something applied. Agriculture is really where DOE and scientific use of statistics (as opposed to betting) was born.

I would definitely do it in the order you have (after all the farming is sort of the economic motivation for the work). Just use numbered subheadings. That's plenty of division. The preceding subheading where you discuss the research topic should be sufficient to explain why you need to cover the two aspects of the problem. (If not make it so...after all, you explained it fine in this question!)

Maybe you are really struggling with what to write about farming, and that is more the issue than the structure. In that case, my advise would be to restrict the topic somewhat to the aspects of farming that were important to your research.