ACM conference papers published as newsletters

zebchuk

There are a few ACM conferences (ISCA and ASPLOS at least, probably others as well) where the conference proceedings are published multiple times: first as part of the original conference proceedings, and then later as part of one or more newsletters.

The common practice (and correct IEEE and ACM style, AFAIK) is to cite the original conference proceedings, including the location of the actual conference (as opposed to the publisher's address).

Unfortunately, Zotero seems to be sometimes combining information from multiple citations, and sometimes simply picking the "wrong" citation. There seem to be a few problems here:

1. Zotero is sometimes combining information from multiple BibTex citations.

2. Zotero seems to be using most of the information from the conference proceedings with the exception of the "booktitle" field. Instead, it's using the "journal" field from one of the other citations.

3. Zotero is using the "address" field instead of the "location" field. I'm actually not sure what the "correct" behavior should be for this case, but common practice from just about every paper I've ever read uses the conference location in citations instead of the publisher's address.

4. Zotero is sometimes ignoring both the "address" and "location" fields.


I think any of the papers from ISCA or ASPLOS will likely have this problem, but I've included two examples below.

In the first example, the information from the first entry is combined with the "journal" field from the second entry. Also, the "address" is used instead of the the "location".

In the second example, there are actually two newsletter publications in addition to the original conference proceedings publication. Zotero seems to be simply using the first entry, except for the "address" field for some reason. The correct behavior would be to use the second entry with its location field.


The first example can be found at: http://portal.acm.org/citation.cfm?id=52400.52432&coll=DL&dl=GUIDECFID=16073284&CFTOKEN=77905982

The corresponding bibtex information is:

@inproceedings{Agarwal:1988:EDS:52400.52432,
author = {Agarwal, A. and Simoni, R. and Hennessy, J. and Horowitz, M.},
title = {An evaluation of directory schemes for cache coherence},
booktitle = {Proceedings of the 15th Annual International Symposium on Computer architecture},
series = {ISCA '88},
year = {1988},
isbn = {0-8186-0861-7},
location = {Honolulu, Hawaii, United States},
pages = {280--298},
numpages = {19},
url = {http://portal.acm.org/citation.cfm?id=52400.52432},
acmid = {52432},
publisher = {IEEE Computer Society Press},
address = {Los Alamitos, CA, USA},
}

@article{Agarwal:1988:EDS:633625.52432,
author = {Agarwal, A. and Simoni, R. and Hennessy, J. and Horowitz, M.},
title = {An evaluation of directory schemes for cache coherence},
journal = {SIGARCH Comput. Archit. News},
issue_date = {May 1988},
volume = {16},
issue = {2},
month = {May},
year = {1988},
issn = {0163-5964},
pages = {280--298},
numpages = {19},
url = {http://doi.acm.org/10.1145/633625.52432},
doi = {http://doi.acm.org/10.1145/633625.52432},
acmid = {52432},
publisher = {ACM},
address = {New York, NY, USA},
}




The second example is http://portal.acm.org/citation.cfm?id=1736021&CFID=16073284&CFTOKEN=77905982

With Bibtex information:

@article{Brewer:2010:TDR:1735971.1736021,
author = {Brewer, Eric A.},
title = {Technology for developing regions: Moore's law is not enough},
journal = {SIGPLAN Not.},
volume = {45},
issue = {3},
month = {March},
year = {2010},
issn = {0362-1340},
pages = {1--2},
numpages = {2},
url = {http://doi.acm.org/10.1145/1735971.1736021},
doi = {http://doi.acm.org/10.1145/1735971.1736021},
acmid = {1736021},
publisher = {ACM},
address = {New York, NY, USA},
keywords = {developing regions, ictd, it for development.},
}

@inproceedings{Brewer:2010:TDR:1736020.1736021,
author = {Brewer, Eric A.},
title = {Technology for developing regions: Moore's law is not enough},
booktitle = {Proceedings of the fifteenth edition of ASPLOS on Architectural support for programming languages and operating systems},
series = {ASPLOS '10},
year = {2010},
isbn = {978-1-60558-839-1},
location = {Pittsburgh, Pennsylvania, USA},
pages = {1--2},
numpages = {2},
url = {http://doi.acm.org/10.1145/1736020.1736021},
doi = {http://doi.acm.org/10.1145/1736020.1736021},
acmid = {1736021},
publisher = {ACM},
address = {New York, NY, USA},
keywords = {developing regions, ictd, it for development.},
}

@article{Brewer:2010:TDR:1735970.1736021,
author = {Brewer, Eric A.},
title = {Technology for developing regions: Moore's law is not enough},
journal = {SIGARCH Comput. Archit. News},
volume = {38},
issue = {1},
month = {March},
year = {2010},
issn = {0163-5964},
pages = {1--2},
numpages = {2},
url = {http://doi.acm.org/10.1145/1735970.1736021},
doi = {http://doi.acm.org/10.1145/1735970.1736021},
acmid = {1736021},
publisher = {ACM},
address = {New York, NY, USA},
keywords = {developing regions, ictd, it for development.},
}

ajlyon

The problem of conference place versus publisher place is due to be resolved in the next release of Zotero-- the issue is discussed here: https://github.com/ajlyon/zotero-bits/issues/6

Furthermore, some types are currently missing a place field-- this will also be addressed in the next major release of Zotero.

The "booktitle" field is being saved as the proceedings title on my computer. Are you sure that this precise citation misses it upon import?

I'm not seeing any evidence of citations being combined-- I get two and three citations for each block above.

zebchuk

I'm glad the place/address issues are being addressed. I look forward to getting the next release when it comes out.

I'm not sure how you're getting multiple citations, when I try to save the item to Zotero, I only get one entry.

For the page at http://portal.acm.org/citation.cfm?id=52400.52432&coll=DL&dl=GUIDECFID=16073284&CFTOKEN=77905982

Zotero ends up with proceedings title "ACM SIGARCH Computer Architecture News" and series = ISCA '10

If you look at the original BibTex entries, you'll see these values appear exclusively in different entries. The proceedings title is coming from the newsletter entry "journal" field, and the "series" field only exists for the conference proceedings entry.

The actual proceedings title should be "Proceedings of the 15th Annual International Symposium on Computer architecture" which is in the "booktitle" field. But that doesn't appear in my Zotero entry at all.

I'm using Zotero 2.1.5 on Firefox 4.0 on Mac OSX 10.6.7. I saw a related forum discussion about conference papers appearing as journal papers, and I've updated my translator to use that fix. That doesn't seem to help with the problem I'm seeing.

For more examples of this behaviour, you can see

http://portal.acm.org/citation.cfm?id=1815963&CFID=16149759&CFTOKEN=25650017

or any of the papers in the Table of Contents at http://portal.acm.org/citation.cfm?id=1815961&picked=prox&cfid=16149759&cftoken=25650017

(not all of these papers have the problem of combining "series" and "journal" fields from different entries, but everyone that I've tried is creating a Zotero where the proceedings titles is "ACM SIGARCH Computer Architecture News" and they should all be "Proceedings of the 15th Annual International Symposium on Computer architecture")

ajlyon

Looks like the translator was being aggressive in overriding data from BibTeX with data scraped directly from the page.

Please go to http://github.com/ajlyon/zotero-bits/raw/master/ACM.js and save the file to the translators directory of your Zotero data directory (http://www.zotero.org/support/zotero_data).

It should start working again. If this works for you, please post here so that I can submit this change to be pushed to all users.

zebchuk

This solves some of the problems. At least it's not ignoring booktitle and it's no longer combining the "journal" field from one entry with the "series" field from the other entry.

But there's still an issue with having multiple citations for the same paper. Zotero seems to be picking the first citation and ignoring any later ones.

For instance, see http://portal.acm.org/citation.cfm?id=1815961.1815990&coll=DL&dl=GUIDE&CFID=16212739&CFTOKEN=45941158

It picks the "SIGARCH Comput. Archit. News" publication instead of the publication in "Proceedings of the 37th annual international symposium on Computer architecture"

I'm not sure if there's a good way to resolve this issue. Maybe use the reference with the lowest doi value? (comparing doi's for different papers, there seems to be part of the doi specific to the paper, and part specific to the publication, and the publication part seems to be chronologically ordered so picking the lower doi value should be the same as picking the earliest publication.)

I don't suppose there's any mechanism that would allow the user to choose which citation was added to Zotero when multiple citations our found for the same item?

ajlyon

This is a very funny situation to be in. It would be possible to present the user with both citations and ask you which one(s) to download-- but that's a fair bit of extra legwork and complexity, with perhaps only a little benefit. Are these articles very common?

zebchuk

This problem exists for the two largest conferences for computer architecture - ISCA and ASPLOS. For my research, and anyone in my field, probably over 50% of the papers I cite are from one of these two conferences.

For anyone doing computer architecture research, this is a significant nuisance. I either have to manually change about 50% of the papers I add to Zotero, or I have to modify any bibliography that I create with it.

Having said that, I'm not sure how often this happens for other conferences in other fields. It might be very rare indeed.

How does this situation compare to what happens when Zotero imports multiple citations from a page of search results, for example. Could the process used there be applied easily to this case?

ajlyon

We could use the same interface of prompting the user to save, but I'm not happy with the prospect of a user running a search and clicking the folder icon, selecting several items from the available ones using the Select Items dialog, then being prompted again to select which publication of each item should be saved-- this would violate the established user experience expectations of one such prompt per save, and it could lead to N+1 such prompts, for N items selected from the initial search results.

It's not feasible to fold in the multiple publications to the original list of items for select items.

It would also be possible to create two items by default from such pages, but that also violates the user expectation that we create one item for each item they select.

zebchuk

How difficult would it be to save the user's preference when this occurs? In my situation, it's always the same newsletter but with different conferences. It really only needs to ask me once and I can tell it that I don't want it to use the newsletter citation.

In the worst case, this might still result in N+1 prompts, but that situation should be exceptionally rare (it should require each item having this problem with different publications).

Given that the current situation is effectively random in which citation gets chosen, how bad would it be to implement a simple heuristic that works for known cases. You could give lower priority to titles with "News" or "Notices" in them, or add a set of hard-coded known instances, or select the lowest valued DOI, or something similar.

Such a heuristic wouldn't change the user experience, and it would always select only one item. It would make it more likely to select the "correct" citation for known cases, and the only down side is that some cases might potentially always select the "wrong" citation instead of randomly selecting one as it does now (I'm assuming the order that ACM uses to list the citations is effectively random).