Home | Research Pages | Pages for Grad Students

Bill Gardner / John's tips for research surveys

Last updated 25/Sep/08

["John" is John Carter , currently a PhD student from Nova Scotia under my supervision. His claim to fame is having worked on a project with NASA Goddard Space Flight Center. He is about the only student I know who actually seems to enjoy searching the literature to find related work. On this page he shares the tricks of the trade that he has accumulated during his years of masters and doctoral work. --BG]

For the sake of discussion, let's assume you're doing a survey on a topic, T.

  • 1. Synonyms. You need to identify all the possible ways that T can be worded.

    • For example, T could be: "integrated widgetry." Variations could include: "hybrid widgetry," "combined widget development," etc. It's very easy to miss an entire branch of the research due to word choice.

  • 2. Clustering. You're going to find a number of papers on topic T. You will discover--and it is helpful for reading/organization--that the work will cluster based on:
  • "Camps": Originator of the work, and their (possibly former) grad students, plus associates
  • Geographic Location: People working on topic T in the US, in China, etc.
  • Approach: People working on the problem in the same way connect their work more often.
  • Time: People doing this 10 years ago, or people who took up the torch for a new-found application.

    • It's important to recognize the clusters for organizational purposes, but also to make sure you're not just looking at the same people, and the same work. It also helps to clarify your thinking, because you'll look for commonality and differences between approaches.

  • 3. Searching. Avoid "meta" searches as much as possible. Instead of searching for T using something like Google Scholar, search each of the following individually:
  • ACM Portal
  • IEEE Explore
  • Citeseer
  • Google (not Google scholar)
  • Google Scholar

    • With meta searches, the most relevant from each source tend to pop up near the top, and get combined into one list. It's way too easy to overlook obscure but valuable work. I liken this to the effect of looking at the books next to the book you've found in the library catalog when you head to the shelf.

  • 4. Look through every page of search results. A lot of the later suggestions will be "misses," but it's easy to check through, and very important stuff can be there.
  • 5. Grab the detailed reference when you find it. Get the bibtex for everything you find. It's much easier than trying to go back and find it later when you've forgotten where you found it.
  • 6. Research on campus is much easier. The sites mentioned in #3 recognize that you're from Guelph (if you're on the UoG network), and provide you with direct access to manuscripts. Some can be done off campus via the library login, but it's a pain, and not a 100% replacement.
  • 7. Perform a search, write a bit, then do the search again. It's an iterative process and insight you've gained from writing will aid in the search process. Avoid the mental trap of:
  • Step 1. Read.
  • Step 2. Write.

    • It never works out that way. I like to think of it as diminishing returns. Search, find about 50% of the material you need. Write what you can. Search, find another 25%. Write more. Search, find 10%. Write. Eventually you'll stop finding anything of any real consequence, and you'll know you're getting close. (But keep looking.)

  • 8. The Library Search Engine is terrible. If you can't find it by title, use author. If that doesn't work, try keyword. Eventually you should get it, but for some reason author, title, keyword can be fairly exclusive of each other...
  • 9. For surveys, the bibliography of a paper is usually more helpful than the papers themselves (intro and conclusion is often enough for the first read). Read the bibliography and highlight everything that looks relevant. Get those papers, and read their bibliographies next.
  • 10. There will always be something you missed, something new or one more thing. Set a goal, and go with your gut in terms of knowing when you've collected enough. Make a list of things you've found but had to exclude.
  • 11. Books. The library has them. Although they're not always cutting edge, poke around in the library catalog. There seems to be a tendency in CS to avoid the library and rely solely on papers. Remember: Books have bibliographies too!

    • Home | Research Pages | Pages for Grad Students