2010Q3 Reports: Publications Chairs

This year, the paper proceedings are not printed due to a tight schedule of the conference. Instead, only the cdrom proceedings are produced. By "cdrom" we actually refer to a card-type USB memory stick. So this is also the first time the ACL conferences use this new media for publishing the proceedings if I understand it correctly. It is convenient for notebooks without a big cdrom drive but with the tiny USB interface. It is also convenient to put it into one's pocket since its size is small. The memory card can also be re-used to hold one's notes taken during the conference. We hope that people will like it.

The nice cover is the artwork of our excellent local committee.

This year, we have both the Linux version of the ACLPUB and a START wrapper of the ACLPUB available. The START wrapper implements the typical ACLPUB make options for producing a single-volume proceedings. Therefore, it is convenient for generating a single-volume proceedings if its input is well-formed, and if the proceedings are not going to be divided into multiple volumes.

Many workshop chairs who are also compiling the proceedings used the START wrapper directly, while some familiar with the Linux version or those who need multiple-volume and cross-workshop integration functions (such as the publication committee and local committee) used the Linux version. Therefore, we have to answer various questions from both sides.

The START administrators, especially Rich Gerber, had been very helpful in answering questions regarding to the START wrapper. They also resolved many problems that had not been considered when the wrapper was initially implemented. We have also established a publication home page to provide various publication instructions, information and some FAQ's so the book chairs can find publication information they may need.

To unify the publication style, we had customized the templates for title page (titlepage.tex), copyright page (copyright.tex) and citation-stamp (citation-stamp.tex) for each workshop, and provided a workshop-specific cdrom meta file to each workshop. Therefore, the book chairs do not need to edit such templates manually and fill in workshop specific information such as the proceedings formal book titles, ISBN's, publisher's information, dates of their workshops, hosting university and city, and so on.

This helps avoid accidental typos and misuse of old templates. It also helps to compile the proceedings quickly. The only requirement is that the publication chairs have to prepare all these templates in advance, and populate them to the START's ACLPUB working environments for individual workshops, or made the customized templates available on the publication home page as early as possible.

Some book chairs are still unawared of such customized templates though, and thus used templates from the ACLPUB ftp site with their own revisions. Therefore, early announcement and contact to the workshop chairs and book chairs are necessary to make it more successful.

We have, however, reports from some book chairs that their templates are not customized in spite of the population of the templates by the START admins. We suspect that the customized templates have earlier time stamps (because they are prepared early) than the default templates, and therefore are replaced automatically by the default templates when the proceedings are generated by the underlying Makefile. This problem and solution will be passed to the START admins so the time stamps of those customized templates be renewed unconditionally before their first use to generate the proceedings.

Another important problem with the START's wrapper was that its Schedule Maker are not designed to directly export the conventional forms of ACLPUB's "order" files, which assign time slots to presented papers or special events. The START wrapper had now provided a function to upload conventional ACLPUB order file. Therefore, this problem is consider resolved.

The START's wrapper does not have a concise step-by-step instruction for producing a proceedings as well.

The inherited limitations of ACLPUB still leave the book chairs and publication chairs some problems to resolve manually.

Some PDF files with compatibility level above 1.4 may not be processed correctly with current ACLPUB. Some equtions in PDF files may become invisible when the PDF files are combined into the proceedings and when thr individual files are tagged with metadata (such authors, paper titles, conference names.) We normally asked the authors to re-compiled their PDF files in a different machine. Most problems will then disappear. Such problems seem to be related to the fonts of the PDF files. May be such problems should be re-examined by the ACLPUB authors.

In addition to the PDF files, the major problem is the non-ASCII input in the title and author fields of the metadata fields.

Some will cause ACLPUB to stop before they are converted into LaTex special symbols in the metadata file or be included to the internal conversion table of some ACLPUB perl scripts. This is particularly true for accents characters, since we have many European authors this year whose names include accents characters. Unfortunately, there are simply too many such characters in know character sets. Therefore, we can simply resolve it one-by-one as new OOV characters (wrt the ACLPUB internal tables) are encountered. An alternative is simply to tell the authors not to use accent characters in the final submission page.

On the request of the local committee, we also wrote a simple script to normalize the capitalization convention of the paper titles. We then asked the book chairs of each workshop to revise some paper titles in their metadata files the first time their proceedings tarball is delivered to us. The script is simple but seems to work fine. So we only removed or manually revised a few titles based on those produced by the normalization script. Maybe the capitalization instructions should be included in the final camera-ready instructions, and the suggestions from the normalization script can be presented to the authors when they submit their final paper.

Since author names are not spelled consistently, we also spent some times to merge authors names in the unified author index, and corrected their counterparts in individual author indexes, paper indexes, program schedules, and BibTex entries of individual workshops. If the ACLPUB or START submission pages are also provided with some simple utilities to detect inconsistent spellings of author names, the generation of the unified author index (as well as individual indexes) will be much more easier. However, cross-workshop consistency cannot be detected easily unless START implements some such cross-workshop checking.

Jing-Shin Chang, Publication Co-Chair, ACL-2010. jshin@csie.ncnu.edu.tw National Chi Nan University, Taiwan.