I stumbled upon a web site that counts conference publications and citations. I asked for their count in Algorithms and Theory conferences, looking at the last 10 years. Here is the result.
http://academic.research.microsoft.com/RankList?entitytype=3&topdomainid=2&subdomainid=1&last=10
This data gives some idea of what it means to have a paper accepted at one of those conferences.
For example I have always thought of CONCUR, COCOON, COCOA as conferences of roughly similar prestige: I have never published there and have never attended one, but I know people who do go there, and the names have so many letters in common that they're almost fused in my mind. Now I see that a paper at CONCUR is much more meaningful than at the other conferences with similar names.
The biggest surprise for me is the contrast between RANDOM and APPROX. I've always thought of them as twin conferences, but actually, RANDOM has much more impact - perhaps because it is better targeted. APPROX is broader and its scope is harder to distinguish from, say, mainstream SODA papers, so it only gets the leftovers after SODA has been served, whereas RANDOM, perhaps, gets some specialized papers that are sometimes a little bit marginal for SODA but that are still very nice papers that will be well cited by people in the area.
If you hit a niche and have the leading conference in a particular subtopic, then your conference can have high impact even if the number of papers and of attendees is relatively small.
Just a remark on this list: ICALP means "International Congress of Mathematicians" for Microsoft, and if I am not mistaken CCC (Computational Complexity) does not appear at all!
ReplyDeleteTo my mind, this means we have to be careful with this list since it may not reflect the reality. I would not be very surprised if your surprise come from the methodology they use rather than from the real difference of impact between conferences.
This strikes me as being as pointless as trying to decide on which journals to send papers to based on impact factors. It's the papers that get cited, not the conferences. Sending your paper to the right conference can help it get cited and sending it to the wrong one can harm its chances of getting cited, but measuring right and wrong by how often other papers in the same conferences have been cited is misguided. Really, the question you should be asking is "are the people who are likely to cite my paper going to see it at this conference?" For this reason, I think that matching the topic of the paper to the interests of the conference's audience is much more important than matching the citeability of the paper to the citeability of the other papers at the same conference.
ReplyDeleteWould you be willing to share the web site? I am curious about the systems conferences I go to, and it would be nice to have such statistics!
ReplyDeleteAnonymous, the link is there (you have to copy-paste because I didn't add the html tag). Once you get to that web page you can simply go to the top drop-down menus and ask for systems conferences.
ReplyDeleteB.: yes, that qualification indicates that this is not perfect...
David, pointless for authors because a paper is what it is, and what it is will determine whether it gets cited (assuming the right people get exposed to it). But not pointless for someone who would be thinking about running a conference, nor for someone on a hiring committee who sees hundreds of applicants in other areas of CS with publication lists that are miles long, and who wants to quickly weed out lists that look impressive but that are all about papers in obscure conferences. With few exceptions, I do not know which conferences outside theory are obscure and which ones are well regarded. This might give at least a hint, for a quick first filter.