Student Session Chairs Report for 2002

ACL-02 Student Research Workshop
Alexander Koller, Gideon Mann


1. Program Committee

The co-chairs of the ACL-02 Student Research Workshop, Alexander
Koller (Saarland University) and Gideon Mann (Johns Hopkins
University) were nominated by this year's general chair, Pierre
Isabelle, and approved by the ACL Executive Committee. Susan McRoy
was appointed by the ACL Executive Committee as Faculty Advisor. The
program committee was formed by the co-chairs and approved by the ACL
Executive Committee. The final program committee consisted of 25
student members and 22 non-student members. Of the 47 reviewers, 19
were from North America, 19 from Europe, 8 from Asia, and 1 from South
America.


2. Submission and Acceptance

We received 29 submissions to the Student Research Workshop. The
submission length was set to six (6) pages using the ACL style files.
Because of the unusually low number of submissions, and a number of
reviewers which we had deliberately selected rather too high, we were
able to get four reviews for most papers. 20 papers were assigned
four reviewers each, 3 papers were assigned five reviewers, and 6
papers were assigned three reviewers. Each submission was reviewed by
at least one student and one non-student reviewer. Reviews were
generally done by e-mail.

We accepted 12 of the papers. One of the accepted papers was
withdrawn because it had been accepted at another conference, and we
accepted an alternative paper instead. Some statistics for the
submissions and acceptance are shown below (Tables 1-3).


3. Presentation Format

The papers will be presented in half-hour slots (organized into three
parallel sessions) allocated from the main conference. Continuing
last year's format, for each accepted paper we have invited two senior
researchers to act as panelists to give detailed feedback to the
student author(s). Each slot will thus start with author
presentation (18mins), followed by panelists feedback/discussion
(7mins) and general Q&A (5mins).


4. Panelists

Currently, we are in the process of recruiting 2 panelists for each
presentation.


5. Funding

As in previous years, the faculty advisor obtained funding for student
travel from NSF. An application was submitted in early December (the
deadline for grants to be funded in time for the ACL in July), using the
current call for papers and historical information about the workshop.
$18.000 was requested and awarded.

The student co-chairs accepted funding applications from all students
attending ACL who wished to request support. To avoid conflicts of
interest, McRoy reviewed the applications and made recommendations for
funding. Of the 12 students attending the workshop, 8 requested funding
and all will receive funding which will pay for most of their expenses.
Funding levels were determined based on distance travelled to the
conference. Because there was money remaining, with the permission
of the NSF program director, McRoy also made recommendations regarding
support for students not involved with the student research workshop.
These were forwarded to the Conference chair and the ACL Treasurer.

6. Problems

The co-chairs found the experience from organizing the workshop
invaluable and rewarding, and we note that some problems of earlier
years have been solved for us -- most notably, scheduling and funding
issues. We would like to thank ACL and the general chair, Pierre
Isabelle, for offering us this opportunity.

6.1. Timing

The co-chairs were selected on December 3, which we found rather
late. The late date forced us to prepare the call for papers and set
up the WWW infrastructure very quickly, and effectively pushed
reviewer selection into the new year. Even so, our call for papers
was distributed two weeks after that of the ESSLLI student session, a
major competitor for student papers, which may have contributed to
the small number of submissions.

6.2. Reviewer Selection

It took us all of January to compile a satisfactory list of
reviewers. By the time we could send invitations, some potential
reviewers had already committed to reviewing for other conferences,
and either declined or asked to be assigned only few reviews.

We spent one week of this period building up an initial proposal, and
three weeks revising it to achieve a better geographical balance of
reviewers. In order to improve efficiency in the future, we advise to
make the requirements with respect to balances in the ACL Conference
Handbook more precise.

To improve the efficiency further, we suggest that reviewers for the
Student Research Workshop should be confirmed by the General Chair,
not the ACL Exec. This would make it possible to get early feedback
from the final decision maker, which is much harder if the final
decision is made by a committee. In addition, this would allow for a
more flexible allocation of reviewers. We chose a very high number of
reviewers, which turned out to be too many for the submissions we had,
because we found it impractical to enter a committee discussion about
additions to the reviewers list.

6.3. Student Status

All submissisions were accompanied by either a CV or a letter from the
student's advisor. This ensured effectively that all submitted papers
were really written only by students. However, what precisely
constitutes valid proof of studentship, and whether undergraduate
students (who don't necessarily have an advisor) are eligible to
submit, was not well defined.

For the future we advise to require that each author should send an
email stating informally that they are a student and aren't planning
to submit their PhD thesis within a year after submission of the
paper. This should have the same effect while removing bureaucracy.

6.4. Distribution of Accepted Papers

Of the 13 accepted papers (including the reserve paper accepted after
another's paper withdrawal), 8 were submitted by students from only
four universities: Two each from the Carnegie Mellon University,
University of Edinburgh, University of Cambridge, and Saarland
University. We made the acceptance decision purely on the grounds of
the reviewers' recommendations, but we are not happy with this
clustering of successful papers to only a few universities: Those are
exactly the places where a student has good access to feedback anyway.

Future acceptance decisions should be monitored for whether there is a
trend towards a focus on a small number of universities, and possibly
discuss measures for reversing it.



Table 1: Papers by country

Country	Submitted	Accepted
USA	11 (37.9%)	5
UK	6 (20.7%)	4
Germany	3 (10.3%)	2
Japan	3 (10.3%)	2
France	2 ( 6.9%)	0
Australia	1 ( 3.4%)	0
Canada	1 ( 3.4%)	0
Turkey	1 ( 3.4%)	0
Vietnam	1 ( 3.4%)	0

"Accepted" counts both the paper that was withdrawn and the paper that
was accepted as an alternative.



Table 2: Papers by geographical area

Area	Submitted	Accepted	Acceptance rate
North America	12	5	41.7%
Europe	11	6	54.5%
Asia	5	2	40.0%
Oceania	1	0	0.0%

Table 3: Papers by topics

Topic	Submitted	Accepted
Machine Translation	5 (17.2%)	1
Generation	3 (10.3%)	2
Information Extraction	3 (10.3%)	1
Lexicon	3 (10.3%)	2
Models of Language	3 (10.3%)	0
Corpus-based Language Modeling	2 ( 6.9%)	0
Morphology	2 ( 6.9%)	1
Syntax	2 ( 6.9%)	2
Dialogue	1 ( 3.4%)	1
Information Retrieval	1 ( 3.4%)	1
Semantics	1 ( 3.4%)	1
Speech recognition and synthesis	1 ( 3.4%)	1
Summarization	1 ( 3.4%)	0
Message and narrative understanding	1 ( 3.4%)	0
Pragmatics	0	0
Phonetics and Phonology	0	0
Question Answering	0	0
Multi-lingual processing	0	0
Language in multi-modal systems	0	0