6

I am asking this question regarding to the documents (thesis, research papers, and publications) that contain original, novel work.

Before starting to write a research paper/ thesis, searching for a problem statement, submitting to a journal, staring research work for Ph.D., etc., every researcher has to do a proper literature search. It is a well-known fact that literature search is mandatory for every professional researcher.

There is a finite number of existing disciplines and each discipline has a finite number of existing topics to research and to publish upon. Why is there no centralized mechanism to document the research that has happened till the previous year? If the discipline or topic is totally new anyway there is no need for the literature search.

If there is such a mechanism, then there is almost no chance for rediscoveries and there is no need to search for the whole literature that is randomly distributed across the Internet, paid journals, libraries, etc.

  • 13
    Are there a finite number of questions? – Solar Mike Sep 1 '18 at 9:38
  • 8
    The point in literature research is not to compile bibliographic information but to READ the relevant papers and understand what the current state of the art is. – Karl Sep 1 '18 at 13:37
  • 2
    @Karl: That's what the question is about. Why don't we have a better way of accessing that information than reading hundreds of individually authored and individually structured, partly overlapping papers? – O. R. Mapper Sep 1 '18 at 13:43
  • 4
    @O.R.Mapper We have that, it's called "textbooks", or "wikipedia". ;-) – Karl Sep 1 '18 at 13:54
  • 1
    How will you know what you don't know without looking at what's known but not already known to you? – J.R. Sep 1 '18 at 23:14
15

These central repositories exists - they are called a (digital) library. What we typically mean with "doing literature review" is going into one or more digital libraries and retrieving all information on the subject. Given how heterogenous research questions, approaches, and scientific results are, it seems fundamentally impossible to provide much more structured information than that across disciplines. That said, for some fields (notably medicine, as I understand it), more fine-grained repository structures have emerged, which allow researchers to query for specific data sets etc.

12

"If the discipline or topic is totally new, anyway there is no need for the literature search." : But how do you know your totally new topic on hypermetabolic quiver systems isn't the same thing as what Prof. X's work called totally supernormal inverted forms?

For example, I am a mathematician. I was recently leading a group of students in studying Paley graphs, which are defined on elements of finite fields. There are all of the obvious papers on Paley graphs to look at, of course; but then there are also papers that concern large families of graphs that may or may not include the Paley graphs. Then there are papers on finite fields that don't mention graph theory at all, but really describe Paley graphs. Or papers in finite geometry that are really talking about Paley graphs using slightly different language.

So unless this repository actually included all of the results of every single paper ever written in full, I'm not sure it would provide much use. Of course, there are only a finite number of results in a finite number of topics that have been published about, so this is theoretically possible. Even then, looking through this repository would be not much different than what is usually called a literature search.

  • The Annual Reviews series of journals have some of the highest impact factors across the board, so I wouldn't go so far as to say "I'm not sure it would provide much use". – Allure Sep 2 '18 at 2:49
  • "But how do you know your totally new topic on hypermetabolic quiver systems isn't the same thing as what Prof. X's work called totally supernormal inverted forms?" - you might not, even with literature search. I'd even consider "same thing published twice with different terminology" as a good example of a flaw in our current approach to literature search. I think the cited claim about new topics can much better be refuted with a question like "But how do you know your totally new topic on hypermetabolic quiver systems accurately represents <something> without measuring it with the ... – O. R. Mapper Sep 3 '18 at 13:52
  • ... methods proven to work based upon <some other discipline>?" – O. R. Mapper Sep 3 '18 at 13:53
  • @Allure Possibly. But that series does not have a title in math (it does have one instatistics though). – Tobias Kildetoft Sep 3 '18 at 16:42
2

I agree literature search in its current form, by reading loads of free-form documents with large overlaps1 to extract the relevant bits and pieces is not efficient.

Let's first look at your individual statements:

  • I agree with the assessment that

    There is a finite number of existing disciplines and each discipline has a finite number of existing topics to research and to publish upon.

    though I consider it quite theoretical: We are probably far from knowing all disciplines that will ever be relevant, thus it's not like a "map" where we could check which areas are still blank.

  • The claim

    If the discipline or topic is totally new anyway there is no need for the literature search.

    strikes me as questionable, though. "Totally new" topics do not emerge out of nowhere, they result from unexpected findings in existing disciplines that pave the way for what can be called a new topic.

Now, a "central repository" across all fields (because, also based upon what I wrote about "new topics", they are all linked in some way) would indeed be helpful, but is not yet feasible for two reasons:

  • Doing and maintaining it would be a tremendous effort. Do not get me wrong, it would be much less of an effort than the net effort of all the researchers around the globe individually reading all those original papers, but as long as there is no good worldwide system to compensate those involved for the effort (and I don't really see one for now), the decentralized approach of everyone just individually publishing their source documents and leaving most of the "linking" of source documents for further use up to everyone on their own2 looks like the only achievable solution.

  • Our technology is not far enough yet. The topic of formalizing knowledge is still in its early stages (when compared to the task of representing complex knowledge on arbitrary topics), and without that, any plan to summarize, juxtapose, and link papers can only be implemented by humans reading natural language text and, at best, producing condensed natural language text that may bring different results into a uniform scheme. Automatization of that process is not yet feasible, leading back to point one.


1: Note that I am referring to the text here, not just the novel findings or results. The results should indeed be overlap-free, but the text describing them is only a part of a paper beside e.g. problem statements or summaries of related work.

2: Except for some comparably localized efforts including survey papers, textbooks, encyclopediae, and similar.

0

There is such a repository - see the Wikipedia article on Annual Reviews.

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.