Exercise 2: Literature Review on a Specific Extended Reality Topic

CS291I – Extended Reality, Spring 2024

Change Log:
Thu., April 11, 3:00pm: posted, exercise announced in class

In this homework exercise, every student individually will do an in-depth literature search and initial review on a specific topic related to the class theme, Extended Reality.

Your topic of choice should (but doesn't have to) be related to your current class project idea. It is quite possible that this exercise will influence or even completely change your class project idea. That is ok, and even expected. Everyone has to do this exercise individually, even if members of an already formed project group do and document literature searches on the same overall topic. Don't compare notes while you are doing this assignment, but as soon as all members have submitted the assignment, pool your resources.

Name and describe your topic

The first part of this exercise is to name and describe the topic that you are researching. Give the topic an overall name, describe it in a paragraph, and list a set of keywords and key phrases that describe the topic. This will be really helpful for your searches. Update this section (especially the topic name and keyword list) in response to successes in finding related literature.

Identify the ten most relevant research papers on your topic you can find

For getting started with your search, look at "Document your search procedure" below. In the end, you want to arrive at (at least) ten of the most related research papers to your topic. "Most related" is not strictly defined. You might have some papers that are not highly cited (yet?) but are in strong topical overlap with your idea, and others that are very impactful, with high citation numbers, but maybe only intersecting with your topic.

For each paper you end up with list all bibliographical information, as well as its citation count. Also, beyond just naming the authors and their affiliations, identify the research groups that were most involved in authoring the paper, and check other papers from that group.

The list of 10 papers you arrive at should be distilled down from a much larger set you examine and consider. Keep all those papers in a spreadsheet or literature review tool (Zotero, Mendeley, EndNote, papersapp.com, BibTeX (& associated managers), ...). Here is an example spreadsheet. The format of your list is not strictly set, but should give you enough information to aid in your search, so identifying title, year, venue, authors, research group(s), #citations, and column(s) for your comments/annotations makes a lot of sense.

At least one or two of your papers should be foundational works

Not all of your papers should be recent works. At least one or two should be pioneering or foundational works. These are typically older papers with thousands, or at least hundreds of references in Google Scholar. I am not giving an exact age range you should look for, as this is very topic specific. It will probably be quite clear if you found a foundational paper or not. For some special-purpose topics, there may not be an close-fit foundational paper yet. In that case, include at least one foundational paper with the best fit you can find. Highlight foundational papers in your result list.

Document your search procedure

Beyond just providing your final list, please document the procedure that led you to your results. We discuss some search techniques in class, including

• Keyword search (Start with Google Scholar, but also try other academic search engines: Semantic Scholar, dblp, Microsoft Academic, Science.gov, CORE, BASE, Baidu Scholar,...)
• AI search (GPT-4 or Bing, be wary of everything, especially GPT-3, Gemini and check if suggested papers actually exist. Also some LLMs may have training date cutoffs).
• Search for survey papers in the area, snowball from there
• Snowballing: Once you found a highly relevant paper, check all it's references
• Reverse snowballing: Once you found a highly relevant paper, check (e.g. in Google Scholar) all the more recent papers that cite this one
• Search by association - after identifying some groups/organizations that do valuable work in the area, look at their other publications
• Hand searching: This involves manually searching through the tables of contents of relevant journals or conference proceedings to identify articles that may be relevant to your research

(Once you find a new tool that really helped you, please post information about it in the #literature-search channel of our Slack.)

The writeup documenting your search procedure should be at least a few paragaraphs long and talk about what approaches you tried, and what worked best for you.

Pick and document your 10 most relevant papers

List the ten papers you arrive at as being most relevant, and explain why you picked these. Maybe they represent different aspects of your topic/field. Maybe you included a good range of older and newer papers. Maybe you included one or more survey papers you specifically liked for more coverage. Argue why this set is helpful for informing research in your chosen topic/field.

Finally, try to summarize the knowledge from these 10 papers succinctly. AI is really good at this. You could ask an AI to distill a summary, or a list of research problems and solutions from the papers (or just from their abstracts) and the output of that may be quite informative to you. Of course, writing your own summary will make you learn even more! Or start with some AI help but form your own summary from those ideas. That's a great skill to pick up!

Submission

Please document all your findings in a writeup (overall, I would expect it to be about 2 papges long, but it can vary), and submit it as a PDF on Canvas. Also submit a spreadsheet with all the papers you considered (see above), and the ten top ones, as well as foundational works highlighted.

Submission deadline is Wednesday, 04/17, 23:59:59