7

I want to write a literature survey on a topic, let's say it's risk of crime due to a new technology, let's say due to the Gutenberg press. However, when I search for the overall theme (crime and "Gutenberg press"), it returns few results. Yet, I know of some papers that fit into the topic that are appropriate, it's just that they each use different terminology (e.g., mass-produced slander or text that undermines the Church's authority). These relevant papers don't link their results to the overall theme, namely crime.

The problem is that to perform an exhaustive and methodological survey is difficult, since it depends on me knowing links between papers I have read and the theme I want to investigate. I was wondering what methodologies I can use to address this problem?

For example, in comparison often a literature survey's methodology section will have something like "we are interested in everything to do with X within topic Y" and base search terms on those two aspects and summarise the results from each database. Then, following sections cover the area based on those results. I would like to achieve something similar, but the ability to search the literature in my case is limited.

  • 2
    Where are you searching? Many scholarly databases have controlled vocabulary to help smooth out the different keywords chosen by individual researchers. Look for something like "Thesaurus" or "Subjects" somewhere on your database page (it's not always self-evident; in PubMed it's "MeSH"). Also, yes, absolutely seek out a librarian at your institution's library. They may be listed as research librarians or subject librarians—either way, they will be happy to help you, and will likely save you a lot of time and aggravation. – 1006a Oct 5 '17 at 4:55
  • Thanks, I had a look. I'm using Scopus, Web of Science, Social Science Research Network (SSRN), and Philpapers. They don't appear to have a thesaurus. The main issue is that I know some relevant papers in my research area, but it's not just that they use different terms, but that these terms are not explicitly linked to the overall theme I am investigating without me putting some thought into it (e.g., a paper may show how the gutenberg press can be used for "mass-produced slander", which we know fits into the theme of "crime", but is not necessarily synonymous with crime). – Dr. Thomas C. King Oct 5 '17 at 9:47
  • Thanks for all of the answers, I've taken up all of the suggestions. I selected Tripartio's response for the cited papers, which provide a methodology. – Dr. Thomas C. King Oct 8 '17 at 18:47
3

First of all, I echo the comments that recommend consulting a research librarian. They are professionals who are experts in finding information like this.

As far as what you can do yourself, I recommend combining two fundamental literature search strategies: synonym searching, and then forward and backward citation searching. Warning ahead: both phases involve a lot of work, since your scenario is non-trivial.

Phase 1: Synonym searching: The idea here is to search for as many synonyms as possible of your two key concepts. Here is the general procedure:

  • List all the synonyms you can for each of the two concepts. In your case, that means to list all the synonyms you can for crime, and all the synonyms you can for Gutenberg press. Using a standard thesaurus (e.g. thesaurus.com) could help here.
  • Search in each database for every possible combination of every synonym. Yes, that means that if you have d databases, and c synonyms for crime and g synonyms for "Gutenberg press", you will do dcg keyword searches. Yes, that's a lot, but it's not over yet.
  • In each article that you uncover that you consider valid, examine the paragraph that defines the concepts (crime and Gutenberg press in your case) and see if these descriptions offer more valid synonyms that you didn't think of initially.
  • Repeat the searches with all possible combinations of any new synonyms that you uncover.
  • After doing all this, if your keyword searches did not uncover any of your initial personal list of articles that you know to be relevant, then carefully try to understand the shortcomings of your keyword searches.
  • Based on your new understanding, revise or extend your keyword list, and repeat the searches.
  • Keep on doing this until you come up with nothing more new.
  • Prune the list to remove duplicate articles.

Phase 2: Forward and backward citation searching: The idea here is to find all the articles that either cite (forward) or have been cited by (backward) all the articles that you have thus far identified. Here is the general procedure:

  • You begin by combining two lists (if you haven't already done so): your initial list of articles that you know to be relevant, and the list of relevant articles you identified from the synonym searching in Phase 1.
  • Backward citation search: Examine the reference list of every article known to be relevant. Ideally, you should look up the abstract of each article; practically, you might only look up articles with suggestive titles, since looking up every single abstract would be very time-consuming. Whenever you find any relevant article, add it to your list of articles to be processed.
  • Repeat this backward citation search for every new article that you identify to be relevant.
  • Continue going backwards in time until you feel that any new article uncovered would be too old to be relevant.
  • Forward citation search: Look up in Google Scholar every article known to be relevant (including the ones added from the backward citation search). Examine every article listed to cite each article. Ideally, you should look up the abstract of each article; practically, you might only look up articles with suggestive titles, since looking up every single abstract might be very time-consuming.
  • Whenever you find a new relevant article, then go back and do a full recursive backward citation search on it, and also do a full recursive forward citation search on it.
  • Continue recursively iterating until no new articles are found.

So that's the crazy, insane way that you could thoroughly do this. Practically speaking, you might run out of steam and simplify some steps, but then your search would not be exhaustive, so that is your scholarly decision to make.

Here are two articles that go into further detail on each of these steps:

Finally, if there is a researcher mailing list that you know whose subscribers study related topics, it might be helpful posting a request there as well.

3

I find searches based on keywords to be quite limited. You can play around with various keywords and see how far you get, but this only helps to get a first cut.

To deepen your search, use a snowball and reverse snowball procedure. This should get you around the problem that authors use different terminology. Be aware, however, that this may introduce some selection bias that you need to hedge against.

One way to reduce the bias and to broaden (rather than deepen) the search in addition to starting with diverse keywords is to ask around. You can just asks your colleagues but you can also get as formal as you want and conduct expert interviews or even surveys.

  • 4
    Perhaps the best thing to do is seek out a research librarian. They have the skills and background to perform better then Google in situations like this (and access to research databases). – Jon Custer Oct 4 '17 at 14:20
  • Thank you for your answer, I am interested in the idea of talking to experts and librarians. To elaborate on the situation a bit more, my overall topic X may not have an expert, but in each sub-topic Y there are experts. The tricky part is finding out which Ys are sub-topics of X, which is where the biases are introduced. For example, if I talk to an expert on sub-topic Y, then my choice of expert seems highly dependent on my narrow knowledge and my ability to link sub-topic Y to overall topic X. Is there a way to manage the biases in this situation? – Dr. Thomas C. King Oct 5 '17 at 9:59

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.