ACL-02 ProgramChairs Report

by Eugene Charniak

Initial actions of the program committee were routine with one
exception, we decided to accept only electronic submissions, a first
for ACL. This worked well due in large part to the abilities of one
of the program chairs, Dekang Lin, who handled this responsibility.
As in previous years we required electronic notification. In
retrospect the usefulness of electronic notification is not completely
clear if one is also doing electronic submission. One seemingly
obvious benefit, allowing the program committee to start divvying up
papers before they were submitted did not work well because of the
large difference between the number of notifications and submissions.
We received approximately 340 notifications of submission but only 258
actual submissions.

Of the 258 submissions one was withdrawn during review, and one was
rejected without review (obviously not relevant to the conference).
The submissions came from 22 countries and by percentages came out as
follows: Americas 130, Asia-Pacific 53, and Europe-Africa 75.

Following the recommendation of last years program chairs, while the
reviewing was blind, the area chairs were given the names of the
authors, to better allow them to detect conflict of interest.

The program committee meeting was held March 30 and 31 in Toronto at
the University, where Graeme Hirst made the local arrangements. (We
did not select the locations of the program chairs because of flight
availability.) The meeting was attended by the two program chairs and
8 of the 9 area chairs. To concentrate discussion on those papers
that were most in doubt we organized the discussion as follows: The
papers were sorted by the reviewer's average "overall" scores. We
then went down the list starting from the top. Any paper with a score
of 4.0 or higher (out of a possible 5) was accepted unless the paper's
area chair had reservations. There was one such case, and if memory
serves, it ended up rejected. Any paper with a score lower than 3.0
was rejected, again unless the area chair had reservations.
(Ultimately all such papers were rejected.) All of the papers in
between were discussed in detail.

Of the 258 papers we accepted 66.

We had many cases of dual submissions. In all cases except one the
authors chose to present at ACL. (The one exception will appear at
this years Cognitive Science conference.) Thus there are 65 papers to
be presented at the main session of the conference. The acceptance
rate overall was 25%. By geography this broke down: Americas 40
(30%), Asia-Pacific 9 (17%) and Europe-Africa 17 (22),

By area this breaks down as follows:
Area Accepted Rejected
Discourse and Dialogue5 21
Generation and Multi-Modality 7 17
Machine Translation and Multilinguality622
Lexicon and Semantics 7 19
Speech, Language Modeling
and Statistical Methods
9 22
Word Segmentation, Shallow Parsing,
Chunking and Tagging
Syntax, Grammars, Morphology and Phonology 516
Parsing 1020
NLP Applications830

As is to be expected many papers could fit into several areas, and
there was some re-shuffling to even up area loads. Even so, it is
clear that NLP Applications received the most papers and and "Syntax..."
the fewest.

Following last year we decided to have another "best paper" award.
The program committee selected 9 papers to be considered. (These were
the only papers to receive two nominations for the award from the
reviewers.) As two of the papers were by area chairs, the remaining
seven area chairs and two program chairs agreed to read 9 papers
within two weeks and at which point we would reduce the group to a
more discussable four. In practice it was quite easy to reduce the
group to four. These four were all very strong, but after some
discussion we finally decided that there was one paper that stood
out, and it was decided to give this paper the award and not to share
it (as was done last year).

Best paper selection was the last content-based action of the
committee. What followed was all administrative, including production
of the proceedings, finding sessions chairs, and producing this