How to find out if my specific idea is really new, or if someone has already worked on it?

Question

I started working on my PhD few months ago and I am trying to solve a problem in a particular topic. The topic is quite specific and many people have already worked on it before. I am stuck at the following two scenarios:

• I would like to extend the topic in some particular direction which I think is interesting and new. How will I know if people have already worked on that 'new extended topic'? I google hoping to find something, but I don't. I read some review papers (but most of them are old) and they don't cover any. Also, I think it is very difficult to read all recent papers related to the topic of research in a hope to land on my specific topic.

• I would like to solve an existing specific topic in a new way, which may be quite simple, but looks efficient. But, I have the feeling that people should have already solved the problem in 'my new way'. How will I find if that has been done? I use google again and land up in quite interesting papers but it still do not solve my problem.

How do you tackle the above situations? I think one way is to do a very extensive literature review of the topic, but how will you achieve that? Review papers? Reading 'all' papers in that topic? I tried to do that right at the beginning, but I realized I can not and should not know everything, but should have a very rough overview of the topic and in depth knowledge in my way of solving the problem.

Edit: My research is in the topic of Computer Vision, in the field of Computer Science. My very specific topic is the topic of marker-less augmented reality.

Answer 1

What I find most critical: figure out the nomenclature. Look for articles about your research question in your own words and create a list of keywords that are frequently used. There is usually no consensus, so you will end up with synonyms.

Once you know the words used by the community in your field to describe the concepts and problems, literature searches get a lot easier. My usual approach is to use a large database of publications (in my field Medline is good, I guess your field has its equivalent) and look for the few keywords I listed. Now, don't read all articles of course, do a selection based on the title first, then on the abstract. There is a lot of redundancy in published papers so you will be able to weed out a majority of the papers at this point.

In some fields, senior researchers sometimes write review articles that list the 'open problems' in the field. Even if they don't give such a list, good review papers give an overview of the state of the art and are a great source of secondary citations.

Of course, discuss the matter with your colleagues and adviser, read prominent blogs and if possible go to (a) good conference(s). Conferences are great to grasp what others do and don't, and they help to make sure you don't reinvent the wheel.

Finally, I think you shouldn't worry too much at this point, here are a few words form my answer to this question Is there any research on the prevalence of academic theft?

The chances that your idea is absolutely original is close to zero. Concepts, ideas and theories are almost always due to an intellectual environment in which many researchers are present. Even theories traditionally attributed to a single individual [...] were the result of inputs from several individuals over several years.

The chances that your state of research, access to resources and previous intellectual work is similar for someone else is very small. Research breakthrough are a result of the right coincidence of opportunity, resources and intellectual maturity.

Answer 2

Among the papers you have found, there will be some that most papers in your research direction would reference, such as foundation papers that define the sub-field and its terminology. Look for recent papers that reference them. Pick out the ones closest to your proposed direction.

Recent doctoral dissertations in your general area can be particularly useful. One of their purposes is to convince the committee that the student is familiar with existing research in the area. They typically have a substantial "related work" section that is, in effect, a small survey paper.

Similarly, select the journals or conferences to which you might submit your papers. Your supervisor should be able to give guidance on selecting journals and conferences. Examine their tables of contents for the last few years, looking for papers related to your direction.

You now have some recent papers that are at least a little bit related to your research direction. Read them, and look at papers they reference, always trying to move in a direction closer to your ideas.

If your field is one with commercial implications, you should also attempt a patent search. However, this is very difficult because patents often use different terminology and seem to me to be designed to be unreadable. It will be a little easier if you use the paper trail to get the names of people working in the area.

Answer 3

First of all, finding 'a state of the art' of a research area will take some time! You can do the following while searching for papers:

Quantity Check: Who did publish more papers/journals for the last 5 years in your field. This is a little tricky, because if your field is more on the theoretical side, then you might even need to go back and check the last 20 to 50 years.

Quality Check: Who did publish more high rank journals. This is pretty strait forward, from the list you made in the first step, check what journals they published, and the ranking of each journals. This step, will definitely give you a clear idea, who is doing the real contribution in your field.

Other factors: There are other common sense factors, you can check on them. For example, who is more involved in the conferences in your field; or who has more of an active research group.

Answer 4

First of all, note that proving a negative is often practically impossible. This includes proving that noone had your idea before.

If you find something, then you prositively know that someone else had the idea already. If you do not find anything, there are several possibilities:

• you're really the first one.
• You didn't do your homework properly.

• It has been done & published, but under some name that you did not find.

  I've made this experience (actually more than once): e.g. thought about a problem, eventually solved it. Knowing the solution lead me to further search terms which in the end lead to literature in a not too closely related field (remote sensing vs. spectroscopic tumor recognition) but even retrospectively, I wouldn't associate the remote sensing name of the method with my description of problem and solution: some key term of the remote sensing nomenclature is used already with (several!) different meanings in my field.

  You may be lucky and find the "unknown" search term mentioned in an introduction section of an article you find via "your" terminology. But I've experienced it twice that came across the "unknown" term by accident - hadn't found a paper giving the link, and even after asking around on conferences for several years whether someone knows (of) a solution for ...

• It may be that the key terms are not really suitable for searching.

  One part of this is the already mentioned overlap in terminology. For the example above, we now have a 4th meaning of "soft" in chemometrics...
  Or, while the precise search terms do not yield results, widening the search may lead to impractically large numbers of papers to wade through (see below).

Point 1 obviously is what you hope for, while point 2 is what you fear. Points 3 and 4 I'd say come under the general risks of research life: shit happens, and IMHO it is an illusion to think it possible to prove the absence of a particular idea in the scientific literature.

So I'd recommend to document that the negative results are not due to your negligence. This may be slightly unusual, but e.g. I told my supervisor that I did not find any previous work for something (report the search terms you tried), but that I'm not absolutely sure it was not done as it was not feasible to read through the details of the >80000 pubmed records that list the "parent" method just to ensure that noone has done what I did before under a totally different name and that I gave up after >100 abstracts hadn't yielded anything.
I'd guess that it is good to have numbers if you follow this line of argument.

---

For the literature search, if your reviews are old, do not forget to check which articles cite these reviews.

Answer 5

have you consulted with the reference librarian at your university library? this person can most effectively help you in finding the appropriate data bases to search and assist you in formulating the proper search query (ies). Google is NOT the answer.