11

Based on @JeffE's suggestion here, I see many CS researchers care mostly about conferences deadlines. Even when they want to read papers, usually they check the recent proceedings of different conferences.

Having a paper in IJCAI or AAAI for example, worth more than publishing in many ISI indexed journals with good reputation. I have no hard evidence for this but being in touch with CS research, I see little discussion about journal publications. why is that? is it good for the spirit of research in the CS field?

  • 1
    I don't form the same conclusion you do from JeffE's comment. Specifically, I don't see where it says that CS researchers care "only" for conferences. More significantly, the comment wasn't directed toward an established CS researcher, but toward an undergraduate student. Walk before you run? – J.R. Jul 2 '13 at 0:46
  • @J.R. I know its a strong only statement and I didn't literally mean it since obviously there are many journals (i.e. TKDE,VLDBJ, JAIR, AI Magazine) has high number of submissions. But taking the general view of many CS researchers I met, they really look after top conferences submission deadlines. – seteropere Jul 2 '13 at 1:51
  • @J.R. I have edited the question to be more specific. – seteropere Jul 2 '13 at 1:56
8

There's a reason CS folks cite when we obsess about conferences. The claim is that the field moves so quickly that conferences are more effective than journal for fast turn-around, and so better reflect the speed of developments.

I think this statement is partly true (conferences do have faster turnaround than CS journals) but misses the point entirely (there's no reason journals CAN'T have faster turn around time).

The real reason is the usual one. We got used to having conferences be the primary source of dissemination, and have no pressing reason to change. Having said that, the arxiv is more and more becoming the first choice of reading material and "hot off the presses" material, so I suspect that your question will become more and more moot as time goes on.

  • 2
    there's no reason journals CAN'T have faster turn around time — Yes, there is. Thorough checking of correctness is incompatible with a short reviewing schedule; checking every detail takes time, and referees are busy. (This argument doesn't apply in subfields of CS where proofs are less relevant, of course. And I'm very well aware that referees take longer than they have to.) – JeffE Jul 2 '13 at 10:15
  • @JeffE: I think there are very few CS papers that an expert cannot check for correctness in, say, 2–3 months (and if you are too busy to do that, just say "no"). – Jukka Suomela Jul 2 '13 at 16:11
  • 1
    For me "faster" = "< 1+ years". I think, as @JukkaSuomela says, that there's no reason journal reviews can't be completed in 2-3 months. Consider also the new reviewing processes for VLDB (more experimental) and ICML (experimental + theoretical) – Suresh Jul 2 '13 at 20:18
3

Just to add some points to @Suresh's answer which already lists the most important reasons (at least in my opinion).

While it is true that brand new results are usually published in conferences due to their speed compared to journals (there's no reasons that journals couldn't be faster, but, as things currently stand, they simply aren't), there are still valuable papers to be found in journals.

The first type of papers I usually read from journals are extended versions of conference papers: once the author gets his idea published fast in a conference, if it is a Really Important Thing, there is nothing stopping him/her to take some time and publish a detailed version in a journal.

The other type is overview papers whose value is not in a vast number of new contributions, but usually are the first ones systematically putting a chunk of knowledge in the same place, and possibly giving a new view of already existing techniques/structures/whatever.

  • This, plus the fact that you can often get away with having less data in a conference paper – it's okay to publish a "preliminary finding." The bar is a little lower, and thus easier to clear. – J.R. Jul 2 '13 at 8:37
  • @J.R. exactly. A paper in the conference might explain the technique, give some implementation tips and be purely theoretical or give a little bit of experimental data (and, label the work as "yours"). Then, in the journal, you would add proofs or formalizations to the technique, expand on any tricks for implementation and/or give a pseudocode, as well as extensive experimental results. – penelope Jul 2 '13 at 8:56

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.