Skip to content

Turn Your Notes into a Literature Review

A literature review is not a list of paper summaries joined together with transition words.

Start from the question, not from the papers

Before organising the review, write down the question it needs to support.

For example:

How have immersive interfaces represented remote robot environments, and what evidence exists that these representations improve task performance or operator experience?

That question tells you which comparisons matter. A paper can be interesting and still not belong in this particular review.

Stop writing one paragraph per paper

This pattern is very common:

Smith did this. Le did this. Zhang did this.

It tells the reader that three papers exist, but not what we learn by considering them together.

Instead, organise around a point:

Existing systems mainly represent the remote environment using video, reconstructed 3D geometry or simplified virtual models. Video preserves visual detail but provides limited viewpoint control, while reconstructed geometry supports free viewpoint movement at the cost of latency and artefacts [citations].

Build themes

Themes depend on your question, but useful ones often include:

  • families of methods or system designs;
  • assumptions about the user, robot or environment;
  • datasets, tasks and evaluation conditions;
  • metrics such as time, accuracy, workload or presence;
  • results that agree or conflict; and
  • limitations that appear repeatedly.

In Obsidian, a theme can begin as a concept note linked to several literature notes. The links and Graph view help you see clusters, but you still need to decide what those clusters mean.

Write synthesis paragraphs

A useful paragraph usually does four things:

  1. makes one clear point;
  2. supports it with more than one relevant source where possible;
  3. explains agreement, difference or limitation; and
  4. connects the point back to your research problem.

A rough writing frame is:

Point: What pattern appears across the literature?
Evidence: Which papers support, qualify or disagree with it?
Comparison: Why do their methods or results differ?
So what: What does this mean for our project?

I'm not suggesting that each paragraph should have exactly these four sentences. The structure is just a check that you are synthesising rather than listing.

Be careful when claiming a gap

"Nobody has done this" is a very large claim. It is also usually surprisingly easy to disprove.

A useful research gap is more specific. For example:

  • an approach has not been evaluated under a relevant real-world condition;
  • existing comparisons use different tasks or metrics, so the results cannot be compared cleanly;
  • a practical constraint is repeatedly assumed away;
  • studies report conflicting results and the reason is unclear; or
  • a common claim rests on limited evidence.

Prefer bounded wording such as:

In the literature identified using these databases and search terms, I did not find a study comparing these representations under the same remote manipulation task.

That tells the reader what you searched and what you did not find without pretending you have read every paper ever published.

Combining two fields is not automatically a research gap

Another very common argument goes something like this:

I read about Field A. I read about Field B. I could not find any papers combining them. Therefore, combining A and B is a novel research direction -- and that is what I will do.

Combining ideas from different fields can absolutely lead to good research. However, the fact that you could not find the combination does not, by itself, explain why anyone needs it.

Novelty is useful, but "nobody has done this before" is not the same as "this is worth doing."

A stronger argument starts with the problem:

  1. What specific limitation exists in Field A?
  2. What capability from Field B might address that limitation?
  3. Why is there a reasonable basis to think it will help?
  4. What new difficulty might the combination introduce?
  5. How will you test whether it is actually better?

For example, this is weak:

I found research on Gaussian Splatting and research on robot teleoperation, but few papers combining them. Therefore, using Gaussian Splatting for teleoperation is a research gap.

A stronger argument would be:

Existing remote-teleoperation interfaces often provide limited viewpoint control or rely on geometric representations that lose visual detail. Gaussian Splatting may offer a more photorealistic, freely viewable representation of the remote environment, but it is unclear whether this improves task performance enough to justify its reconstruction time, rendering requirements and visual artefacts. A controlled comparison with video or point-cloud representations could test that trade-off.

Keep the review alive

Your literature review will change as the project changes. Keep useful search queries, record when you last searched and update the Zotero collection when a new theme appears.

Do not wait until you have "finished reading" to begin writing. Write small comparison paragraphs while the papers are fresh. Those paragraphs will reveal missing evidence much earlier than a folder full of completed notes.

Note

For a capstone project, the first literature review does not need to map an entire field. It needs to show that you understand the work most relevant to your question and can explain why your project follows from it.