What to do if a work is too extensive to be properly presented in a single paper?

Question

I have been working for years (started already as an undergraduate student) on a topic in computer science (robotics/AI). The work consists of many sub-contributions (i.e. novel ideas) in several sub-fields. Although each contribution is rather "small" they each play together to result in an over-all method that significantly advances the state of the art for solving an important class of problems.

I'm currently trying to write a paper about it, but it turned out very difficult to "divide" the whole topic into "paper-fitting" pieces. The main concept only works as an "interplay" of the different pieces. Actually this would be ideal for a traditional monolithic PhD thesis but this is not possible because nowadays (in my field and institute at least) PhD theses have to be an "agglomeration" of single papers (which, IMO, is a pity, but that's a different topic).

My current attempt at a "minimal" paper already has 30+ pages and I'm afraid that it is still much too short to adequately cover all sub-topics. For example, in the "result" section alone I would need to demonstrate several sets of examples for each of the 5 main sub-contributions.

So can I best handle this situation?

Publishing several smaller papers seems problematic for two reasons:

1. It's unclear how to "connect" the different papers. Each paper about contribution C_i would have to reference other sub-contributions C_j (j!=i) with a sentence like "here we adopt a novel scheme C_j which will be demonstrated in our still unpublished paper P_j". But obviously, one cannot publish all C_j at the same time.

2. I'm afraid that each contribution as "standing for itself alone" could be considered too "small" or "trivial" by reviewers, especially in high prestige journals. The "interplay" of the sub-contributions is in itself perhaps the most important contribution.

Answer 1

Write the full story, explained as well and concise as you can manage. This will take however many pages it takes. There might be a journal suitable for publishing it, there might not be. In either case, you can make this full story write-up available on the arXiv.

It is unlikely that you'll get too much attention with the full writeup. So you'll then write shorter "advertisement" papers, maybe as conference contributions. These highlight particular aspects, and point to the full story paper for context.

Answer 2

Do you have a supervisor? What do they say? This depends a lot on the content of the papers.

Can you write a first paper that's more of a literature review laying out the class of problems, the current state of the art, and issues with the current state of the art? This can lead into your technical papers which cite your literature review for background.

Maybe the work isn't suitable to be published in a high prestige journal, and your best option is smaller papers laying out the smaller contributions. If your view is that the main contribution is the interplay between these smaller pieces of work, maybe once you've published the foundational work you will have enough to say about the interplay between them for a final paper laying this aspect of it out.

Answer 3

Put the bare minimum into the paper itself, everything else into Supporting Information.

This is the standard trick at least in physics/chemistry and as Alex bGoode commented, it should also apply to AI/robotics. If, for example, explaining a method in depth takes three pages, put a single-paragraph "elevator pitch" into the paper and "see the SI (or Appendix, or whatever your field calls that) for details". Same with proofs, or results that aren't entirely ground-breaking but are just there to confirm that the method works correctly. If you can produce one plot that really "sells" your method, showing just how much better it is than the state of the art, by all means put it straight into the paper. But all the remaining ones that just validate that the method makes sense or confirm that there aren't some nasty hidden surprises in it, those can perfectly well go into the SI.

Basically, whatever you would put onto your slides for a 20-minute talk about the project, goes into the paper. All the rest is SI material.

Answer 4

You say that each of your "pieces" as you name them is probably too trivial for a paper in its own. But then why you need 30 pages in your writeup?

You could simply skip describing your pieces in detail and focus only on the end-result as the cobination of each of the pieces (This is completely fine I would say and many research papers are like this). So just list the "pieces" briefly without going into detail and then describe the result?

If that does not work (and you indeed need to describe details of your pieces because the paper would otherwise be non-understandable) you can put them into an appendinx, or even supplemental material.

Personally I have done this too and I like it when other papers do it. It provides good focus on the main result without omitting the details for the ones who want to reproduce or continue your research.

It is also a good opportunity to learn what is really important and interesting for others. (Much less than one thinks initially. At least I learned that. But keeping it in an appendix or so prevents you from throwing away your research/writeup entirely and actually provide it to those who might find it useful).

Answer 5

Have you considered to write a whole book on the subject instead of a series of paper?

That would give you as much space as you need to elaborate.

Answer 6

Where do you think to publish your paper? Did you evaluate, already, HTML pages connected with links as a solution.

Anyway I would focus on the first page where I would try to highlight the benefits that can be gained from your work, so as to attract the attention of your potential stakeholders.

In this introductory document you could also offer details of your work to those who email you, or by posting all the insights on the Internet.