10

In some research area in CS we got some very interesting results. Now we are thinking about publishing them. In the group we are, the philosophy is to publish inmediately small things in conference papers, which is ok but not the best. Now I am thinking about collecting more of these "small things" and publish in a JCR paper, impact factor more than 2. My question here is how would you decide when you have enough material to prepare a decent paper in a JCR journal? And furthermore, once you think this is ok, how would you decide to which journal you submit the paper to?

PS: JCR refers to Journal Citation Reports, an extensive list that collects info about the most representative and relevant journals in all scientific areas along with their impact factors, so they are usually refereed and no conference papers there

Sadeq Dousti
  • 16,479
  • 9
  • 69
  • 152
Open the way
  • 321
  • 2
  • 11
  • 3
    This surely depends on your exact research area, but it seems to me that in theoretical computer science you would be expected to first publish (parts of) your results in a top conference first, and only later submit to a journal. You might think this is not optimal, and that it can be hard to extract an extended abstract from a full paper, but on the other hand it also lets you reflect on the results a bit longer and gather more returns from conference reviewers and attendees. Telling in which research area you are will probably get you better answers. – Sylvain May 27 '11 at 18:21
  • One criteria could be that you have a well-rounded theoretical account plus appropriate experimental results and perhaps user experience statistics or other such evaluation. Of course, it depends upon the field. – Dave Clarke May 27 '11 at 18:39
  • 4
    What is "a JCR paper/journal"? – Jeffε May 27 '11 at 19:59
  • JCR=Journal of Citation Reports; Journals that appear on this list are usually related with high impact factors. In most countries, for applications, the final score depends on the numerical value of the sum of impact factors, thus important – Open the way May 27 '11 at 21:17
  • BTW, why the downvote? – Open the way May 27 '11 at 21:21
  • FYI my background is physicist and working on the implementation of simulation methods on GPUs and OpenMP/MPI – Open the way May 27 '11 at 21:22
  • @gpu: I downvoted because the question was unclear (see JɛffE's comment). – Jukka Suomela May 27 '11 at 21:28
  • ok, so now is more clear than before right? – Open the way May 27 '11 at 21:34
  • @gpu: Why don't you edit the question to make it more clear? Now you are forcing people to read not only your question but also all comments before they can understand it. – Jukka Suomela May 27 '11 at 21:38
  • you are right! let's go – Open the way May 27 '11 at 21:40
  • 8
    By the way, you have already got a very good answer in the form of @JɛffE's question. In TCS, people don't know JCR; in general, they do not pay attention to JCR or impact factors. – Jukka Suomela May 27 '11 at 21:57
  • I can not understand that. In some countries, like germany and when you apply for funding, the thing that counts the most from the CV is impact factor – Open the way May 27 '11 at 22:25
  • 4
    In computer science, people care more about the reputation of the conference. Certain conferences in each subfield are known to be reputable (and of course certain journals); usually (but not necessarily) affiliated with the ACM or IEEE. –  May 28 '11 at 04:32
  • 9
    Publishing in a high impact factor journal means "I published in a journal where some other papers were well cited". What is the point of that? If you want to tell whether your papers have impact, look at their own citations. – David Eppstein May 28 '11 at 05:29
  • You policy might also depend on your circumstances. In Europe, we do not have as many "big" conferences, as far as I know, and crossing the atlantic all the time is (too) expensive. So I guess journal publication is more common here. Also, journal papers are often (judged to be) better refereed and more self-contained. – Raphael May 28 '11 at 08:32
  • yes, I agree with Raphael – Open the way May 28 '11 at 09:07
  • 3
    @gpu: Regarding funding applications. Usually, you should have computer scientists evaluating computer science applications. And they do know how our field works; in particular, they know [how to find out] which publication forums look impressive in a CV and which are somewhat dubious. And they can also find out how much your work has been cited by others, etc. – Jukka Suomela May 28 '11 at 09:33

5 Answers5

16

JCR is part of the ISI Science Citation Index family of databases, which is not especially reliable in its measurement of computer science impact. See for instance this report from Informatics Europe which specifically warns against using ISI databases to measure the impact of computer scientists.

In any case, if you want your papers to get attention from other computer scientists, publication in selective conferences is still the way to go. We can argue whether it should be, but that's what it is. Some conferences (e.g. SIGGRAPH) count as a journal directly (their proceedings are published as a special issue of a journal), others (e.g. FOCS/STOC/SODA) expect their authors to later turn the same material into an expanded and more polished journal paper, but the conference is to a large extent how to signal other researchers that your work is worth paying attention to. If you're not familiar with the conferences in the area, choosing one sponsored by ACM or IEEE over one that isn't is a good rule of thumb, as is choosing one that calls itself a "symposium" over one that calls itself a "workshop".

However, that doesn't mean that you have to partition your work into least publishable units as you describe your coworkers doing, and it also doesn't mean that saving up multiple small results into a single paper is a good way of making your papers stronger. It's certainly possible to publish papers like that but to my mind the better ones are often the ones that have a single strong main result, perhaps with additional results on the side not to strengthen the paper but rather to provide more complete coverage of its subject.

David Eppstein
  • 50,990
  • 3
  • 170
  • 278
13

Write a paper that you're proud of and that your target audience will be eager to read. (This requires knowing what your target audience cares about.)

There are roughly three classes of journals:

  • Elite journals. These are the rare venues that make people stand up and take notice. There's roughly one in every major field, but remember that a journal that is elite for one field may be considered marginal by others. For example, the rare computer science papers in Science and Nature are usually crap (cough D-Wave cough). If such a category even exists in computer science, it essentially means JACM and nothing else (except in graphics, where it means SIGGRAPH and nothing else).

  • Good journals. As a good rule of thumb, these are the journals that the top people in your target audience publishes in, and that publish the papers that the top people in your field actually cite. Again, which journals are "good" depends on your field. I suppose this is what "JCR" is a proxy for.

  • Write-only journals. Thousands of pages, costing thousands of dollars, that almost nobody reads or cites.

Finer distinctions between these categories are pointless. Other metrics like the number of issues/articles per year, are utterly irrelevant. A long time to publication is frustrating, but also ultimately irrelevant; this is only an issue of latency, not throughput. Assume it will take a year and plan accordingly. Meanwhile, you're already published your results in a good conference, right? And you've put a full version of your paper on the web, right?

Ultimately, your work is not going to be judged on the venue in which it's published, but rather on the impact it has on the research community. There are crap papers in JACM, and there are groundbreaking papers in write-only journals.

So where should you aim your paper? Be respectful but brutally honest with yourself. If you really have a once-in-a-lifetime breakthrough result, send it to an elite journal. If you have a good paper that will be interesting to a large segment of some intellectual community, send it to a good journal. If you have some decent results that could be published somewhere but only a handful of people would care about, I suppose you could send it to a write-only journal, but why bother?

In other words, stop worrying about how to play the game and just do good science.

Jeffε
  • 23,129
  • 10
  • 96
  • 163
1

Here you are a few guidelines I follow.

Pick an interesting problem. If it has been already solved, find the current state of the art solution and try to improve it. If you are able to improve it, then you can publish it unless the improvement is really trivial. If the problem is open, try to find a solution and if you come up with a clever one, then you can publish it.

It is not a question related to how much material you have; rather it is strictly related to its quality.

In order to find the proper journal, you must know the relevant scientific community. People with the same scientific interests usually (with some exceptions) publish their results in the same set of journals. Otherwise, hunt for a set of papers dealing with a topic close to the one you are investigating, and discover where they have been published.

Of course, once you know potential journal candidates, you decide on the basis of additional factors. These include impact factor, number of issues per year and total number of articles published per year, etc. I always try to balance impact factor with a decent number of articles published per year. Another factor I take into account is the average time from initial submission to actual publishing of the paper. Unfortunately, there are not official data available for this, and you will have to estimate this by yourself by downloading some papers and averaging this data.

Massimo Cafaro
  • 2,216
  • 20
  • 19
0

How to decide when you have enough research results to write a journal paper:

As soon as you have matching upper and lower bounds.

Jukka Suomela
  • 11,500
  • 2
  • 53
  • 116
0

Then how do you define not enough research results for publication? I'd advice to talk to your supervisor (if you have one) or Just write your papers (you will feel which area is weak and which area is not) Most importantly, you can then submit to conferences/journals to see what comments there are for you to make your experiment/algorithm better.

Who else could say they have enough research results until they see the word 'accepted' ?

A big part of scientific research is about doing, trying, failing, and accepting.

Gary Tsui
  • 153
  • 5