Starting Your Research

After supervising many postgraduate and undergraduate research projects, and repeating the same set of guidelines and tips for each of my new students on starting a research project, I thought of documenting those. While these are the tips that I give to my students, I hope it will be useful to other students as well. However, make sure to check with your supervisor as he/she may use a different approach while starting research.

Identifying a Research Problem

Identifying and understanding a research problem to work on is the hardest part of any research project. Rather than inventing a new hypothetical problem, look for a problem that already exists around you. Look for a felt need to improve an unsatisfactory solution, improve its efficiency or scalability, or even come up with an alternative solution/process to address the problem in a better way. Look for problems around your house, neighborhood, company, university, social space, volunteer organizations, or even while you commute. Do a quick online search about the problem and potential solutions. If people are talking about it, it means the problem is relevant. Don’t get discouraged, if solutions are already available for that problem, you can always come up with a better solution. Also, discuss with people about the problem(s) you see and get them to talk about the problems they see. Get their feedback about the problem(s) you are thinking about because others may have different views about that problem and may open your eyes to aspects that you didn’t think about. While this means fail fast for many of your ideas, it also means you are learning fast about the area of interst.  A good problem should be relevant, worth solving, and manageable.

I encourage you to come up with a couple of potential problems before the initial discussion with me. While we may or may not work on those problems, it gets you thinking about finding a problem (a felt need to do something) around you. The ability to see/identify problems is an important skill taught by research. For example, if you can identify problems in a product developed by your company and devise simple solutions to those problems, your value as a team player increases (which could even be translated into monetary benefits). Potential problems you describe can also help me to identify what areas you really like to work on. At the end of our initial discussion, most of your problems may get dropped due to various reasons. Some of the reasons include practicality, lack of resources, too theoretical, having no research component, and may not match program, your, or my interests. During the discussion, I may also suggest project ideas in line with your problems. After several rounds of further discussions and a quick literature survey, you will pick a specific area to work on. Eventually, you will end up with modified versions of 1-2 problems that are in line with your and my interests.

Depending on your research program/level, your problem needs a different mix of Research (R) and Development (D). An undergraduate project usually has a smaller R and a larger D. While an MSc project usually has a balanced R and D, a PhD project must have a larger R and a smaller D.

Here are some resources on finding a research problem:

Conducting a Literature Survey

Once you identify an area of interest and 1 or 2 potential research problems, it is important to read openly and widely about the area, problems, and existing solutions. This phase is called the literature survey. A literature survey will help you understand the state of the art, the gravity of the problems, various parameters involved, and possible solution approaches. It can also save you from the embarrassment of repeating existing work without knowing they exist. For example, you don’t want one of the audiences in your thesis defense or conference presentation to tell you that similar research was already completed several years ago!

Initially read widely from any potential sources, including conference/journal papers, white papers, industry articles, Wikis, blogs, etc. Keep track of the literature that you read including title, authors, where it appeared, and URL. You may highlight and write notes on research papers (on soft or hard copies), annotate PDfs (web pages may be saved as PDFs and then annotated), and/or write a synopsis in a separate journal or an online document. Post-it is not a good idea as they may not survive your long research endeavor.

As you develop the necessary background, narrow down your literature search to specific topics and problems you want to work on. Look for patterns/themes emerging from the literature you come across. Pick those patterns/themes to focus your research problem. As you understand more and more literature, your research problem can be refined. In certain cases, you may even completely change the research problem.

During this phase, an undergraduate project needs to cover at least 8+ key literature sources. An MSc student needs to cover 15+ key papers. A PhD student may need to cover 30+ key publications. Following are some resources on conducting a literature survey:

Formulating Problem Statement

Once you are a little bit clear about the research problem, it is important to start documenting it. While most students think that they know the problem they are working on, they get stuck once they are asked to explain it to someone else or write it down. This is a good indication that you really don’t know your research problem. If you really know the research problem, 50% of your research is complete, as it’s much easier to work on a well-defined problem than a problem that is vague. While it is hard to clearly define your research problem, sometimes even at a much later stage in your research, it is important that you attempt to define it early in your research. It’s natural for your problem statement to evolve/change during research as you better understand your problem and limitations.

It is extremely important for you to be able to clearly formulate and specify your research problem (a.k.a problem statement). The research problem should address the following 3 questions in any suitable order:

  • “What” is your problem?
  • “Why” is it important?
  • “How” are you planning to address the problem?

While these 3 questions come from a different domain (see Start With Why – Simon Sinek TED talk), it is relevant to research too.

Initially, try to come with a 1/2 page write up addressing the above 3 questions. Usually, it is easier to start with why. A paragraph format is preferred. If you find it hard to write down your research problem, this is an indication that you are still not clear about the problem. Then you need to discuss with me the difficulties you are facing. It is also a good practice to try to explain your problem to some of your colleagues who may not have much of an idea about what you are planning to do. You need to be able to explain the problem to someone with a basic background in Computer Science, but not an expert. You may also need to look into more literature to better refine your problem.

Once the problem statement is more clear, you need to write a detailed version of it within ~300 words. This is usually Chapter 1 or 3 of your project proposal or thesis. If it is for the project proposal or literature survey, it’s usually relatively short and given as part of Chapter 1. A thesis usually contains a detailed version of the problem statement in Chapter 3. Your detailed problem statement should contain the following:

  • Background/Motivation – Provide sufficient background to the problem and discuss why is it important. Also, briefly discuss a few key use cases, if relevant. Use this as a motivation to your research problem.
  • Problem Statement – A clear statement of the problem you plan to address. Clearly define what the problem is. While most real-world problems are complex and large, usually a research project addresses only a well-defined sub-problem within the main problem. Therefore, clearly state the scope of planned work. State any assumptions.
  • Research Gap – Given the current body of knowledge related to the problem and solving it (as per your literature survey) clearly state what is missing. This is what worth your problem solving
  • Proposed Solution – Briefly outline your solution with specific goals and objectives (i.e., what). Justify why you think your solution will work. Also, discuss how each of those objectives is to be achieved. While you still haven’t started the research, you need to have some idea about how you plan to tackle the problem. The process of tackling your problem is called the research methodology. Present high-level diagrams that illustrate the solution and main components. It’s also a good idea to illustrate your research methodology using a diagram.
  • Research question – Try to come up with a single sentence question that reflects the problem you are trying to solve. The research question usually start like “What is …” or “How to …”.

Sometimes the problem statement is written as 2 steps, namely problem statement and purpose statement. You may write them together or separately based on what’s appropriate to your degree program, type of study, supervisor’s preference, etc. Here are some useful resources:

Writing Literature Survey

The literature review is an endless process. It grows like a tree where a couple of key papers can point to many relevant-looking papers, these papers will point to lots more papers, and so on. At some point, you have to freeze looking for further literature. This can be done once you have a broader and deeper understanding of the area of interest and your research problem becomes clear. Usually, I will give you a hint on to slow down the literature review. In practice, a literature review will end only when you complete the final version of your thesis. You may need to look for further literature during the following phases:

  • Solution design – Typically a solution has many components that you may not be familiar with. For example, you may have to look for multiple alternatives before deciding on what algorithm, database, tool, middleware, or survey technique to use.
  • Solution implementation – During the implementation of your proposed solution, simulation, or survey you need to find out configuration/simulation parameters or better ways to formulate your survey questions.
  • Writing the thesis – When you start writing the thesis you need additional literature to build better arguments for your problem statement and solution. Also, additional literature is usually required to fill the gaps while writing a clear and comprehensive story.
  • Long-term research – If your research spans several years, e.g., for PhD and part-time students, it is important to keep track of new papers that get published while you continue the work. Subscribing to relevant websites, following new papers from relevant journals/conferences, and Google News or Scholar alerts are some of the ways to stay informed.

Once you have read and understood most of the relevant literature, it is important to document them. This document is usually going to be Chapter 2 of your thesis. While preparing this chapter, it is important to critically evaluate the literature while covering all key papers (a.k.a., seminal papers), alternative designs, emerging patterns, and a comparison among them. Make sure to write the summaries in your own words. This also helps you to reduce your similarity score to avoid potential plagiarism. If you find it hard to do so and constantly looking at the original paper for sentences, it means you have not clearly understood the ideas in that paper. In that case, it is better to stop writing and go back and re-read the important sections of that paper (this is where highlighting and annotating papers becomes useful). Otherwise, your write up will not tell a clear, concise, and cohesive story.

Your litetaure review should contain the following:

  • Summary of all related solutions, technologies, algorithms, and tools. Even if your solution is specific to a particular technology or tool, you need to explain other related ones. Then you need to justify why you decide to work on the selected technology/tool but not on others.
  • Use the literature survey to motivate and justify your problem.
  • Include figures and tables whenever relevant. Avoid taking sceenshts rom original sources, instead, comeup with your own.
  • Critically evaluate each technology, algorithm, and tool. Discuss their pros and cons, credibility of assumptions, missing elements, performance results, and why each solution is applicable or not applicable to your particular problem.
  • Summary of each solution and a comparison between comparable solutions. Better to present this as a table.
  • Contents need to be related, connected, and have a good flow. Start with topics that can be easily explained. Then present topics that have dependencies to other topics that were presented earlier. For example, if your project is on processing big data from Internet of Things (IoT) applications, before presenting similar solutions first present what is big data and what is IoT.
  • Properly cite literature.

Before starting to write the literature review, first come up with an outline. This outline should contain the list of topics that you plan to write about. Those topics need to be ordered based on their importance and connection to each other. This helps you and me to identify missing topics, and make sure your write up has a good flow. This is less time consuming and error-prone than moving topics around once you have written everything. At this stage you may also identify additional literature you may have to look into.

While writing the literature survey it is important to use a suitable language and style. The academic writing style is preferred as it emphasizes a clear and concise expression of ideas and findings. It is also important to use an appropriate referencing style while referring to the literature. IEEE, ACM, or APA referencing style is recommended depending on your programme. Here are some resources on academic writing and writing a literature survey: