2012Q3 Reports: Publications Chairs

Summary

On the whole, the production of the proceedings went reasonably smoothly, though the process could be improved in many ways. In particular, the production of the unified proceedings for the USB (17 ACL volumes and 2 EMNLP-CoNLL volumes) was more difficult than expected, and support for supplemental files turned out to be nonexistent. We recommend that ACL invest in improving START to handle both issues, as well as implementing the additional enhancements identified below. Naturally the documentation should also be improved.

Details

• Rare accented characters (eg from Latvian names) as well as latex special characters (eg underscore) were a hassle to accommodate. Over time, the problems should diminish as START is improved to handle these cases and as latex support for unicode improves. The documentation should be improved to state that such character issues should be addressed in START, rather than attempting workarounds with latex escapes.

• It was a surprise to discover late in the process that putting together the USB contents in a way that nicely integrated all 17 ACL volumes (with a unified author index and bib file), as well as 2 EMNLP-CoNLL volumes, involved a great deal of manual effort. Thankfully, Nizar Habash shared the scripts he created for doing this for NAACL, which made the job easier. Nevertheless, besides downloading 17 separate proceedings tarballs (multiple times as they were updated), it was necessary to rename them, extract the contents, rename the extracted directory, then run and debug the script to generate the unified proceedings. Problems with accented characters again arose in alphabetizing the author names and in generating HTML pages, one reason that it would be of great benefit to incorporate the scripts for producing a unified proceedings across volumes into START -- the other main reason being that it would be a huge improvement to avoid all the downloading and renaming of proceedings tarballs for the various volumes. Naturally, the process should also be documented.

• Relatedly, Nizar Habash shared a script for extracting all the program information from the various volumes necessary for Uli Germann's script for generating a handbook for the conference; these scripts were not quite in sync, and Nizar's script needed to be improved somewhat. Having this script in START would again be very useful.

• It was also a surprise to discover that there has never been support in aclpub for including supplemental files on the cdrom or USB; this was discovered too late for this year, though we made concrete suggestions to Rich Gerber for how to do this in START for future years. As it turns out, it's also a hassle to get the supplemental files into the ACL anthology, requiring the START paper ID to anthology paper ID mapping to be reconstructed. Thankfully, solving the USB export problem should also solve the anthology import problem, another reason why it is import to support the development of this capability in START.

• Generally it seems that incorporating aclpub into START has been a plus, especially for the book chairs, most of whom managed quite well with it. However, it does take more time to wait when uploading, testing files and generating even small PDF files (such as title page). Note that there should be a better process for informing in advance all book chairs -- not only the workshop organizers, but also the ACL organizers such as the demos and tutorials chairs -- of their expected duties. The documentation should also be streamlined for single volumes.

• Re copyright forms, for authors who forget to fill out the copyright release, it should be possible to selectively re-enable the final submission forms after the deadline; this would also make it possible for authors to upload corrected submissions when problems are found. (We have suggested this enhancement to softconf.) Additionally, for those with non-standard copyright releases, it should be possible for them to upload these forms for checking later.

• We ran into a problem where authors using Microsoft Word uploaded PDFs with margin problems that caused the copyright stamp process to fail. Fixing these problems required a great deal of time coordinating with authors. It would be much better if the final submission process *required* authors to check and fix their submissions.

• One tricky issue is that pub chairs are often asked to agree to tighter schedules without having any idea what is realistic. This year, for both the workshops and for EMNLP, requests were made to compress the production schedule for the proceedings. It is difficult to know what is realistic when serving as pub chair for the first time, and the lack of printed proceedings makes it seem that the schedule can be compressed. In responding to these requests, it was helpful to consult with the NAACL pub chairs (though it seems that the NAACL pub chairs were a bit ambitious for comfort); in the end, the lack of printed proceedings does not seem to save as much time as one might think, given the time it takes to communicate with authors, book chairs and other conference chairs. Ideally this point should make its way into improved documentation for pub chairs.

• Finally, we would like to note that given the workload, we found it essential to have co-chairs, i.e. having just a single publications chair would not have been feasible. Splitting duties on an ad hoc basis based on availability and expertise worked fine.

---

Maggie Li and Michael White

--MichaelWhite 19:31, 22 June 2012 (UTC)