Summary Report of the General Chair HLT/NAACL 2004
Julia Hirschberg
June 2004


High Points:  

∑ Conference registration numbers are good.

∑ We chose three chairs for each sub-chair position, one each from
NLP, IR, and the Speech community.  This worked extremely well, since
there was always backup when someone was traveling, and it spread the
workload out.  I think it also gave more people a sense of involvement
in the conference; e.g., 21 of the 27 chairs are attending the
conference.  All of the groups did first-rate jobs.  Christy and I are
having a thank-you lunch to show our appreciation on Monday.

∑ The paper submission and selection process went relatively smoothly
and the PC chairs and area chairs all recommend the Start review
software they used to other ACL conferences.  Long paper decisions
were made in a face-to-face meeting and short paper/poster decisions
on a conference call.  The PC chairs chose area chairs who had
expertise in multiple areas, to provide flexibility in case
submissions did not follow the previous year’s pattern, and this
worked well in general.  The PC chairs and Area chairs decided to give
a best paper award; in consultation with the NAACL board, we decided
to do this, although this decision need not bind future PC committees.
43 of 168 full paper submissions were accepted (26%), and 40 of 84
short papers (48%).  While it was difficult to count the submissions
by broad area, the PCs best effort at doing this (which counts some
papers in multiple categories) indicates that, for Long Papers, 129
were in NLP, 53 in IR, and 33 in Speech; for Short Papers, 47 were in
NLP, 28 in IR, and 27 in Speech.

∑ Student Workshop: We did receive NSF funding ($20,164) this year to
support the workshop, thanks to the efforts of the Faculty Advisors to
the workshop.  One issue here is international students, since NSF is
better able to support U.S. students from U.S.; however, other options
for support are available for foreign students, and future Advisors
should explore them.  We also decided to hold the workshop during the
tutorial day, so that there would be less competition from parallel
paper sessions.  Finally, we decided to hold a student evening party
instead of a lunch; IBM agreed to sponsor.

∑ Tutorials: We settled on six, 2 in each of our theme areas, IR, NLP,
Speech.  As of the end of pre-registration, two workshops (one NLP and
one Speech/IR) had rather low enrollments (12 and 8) but the rest were
doing fine.  We decided not to cancel any assuming we would get
walk-ins in Boston.

∑ Workshops: We received 11 proposals and accepted 10.  As of 26
April, registration ranges from 19 (WS7 on Speech Indexing and
Retrieval) to 60 (WS8 on Linking Biological Literature, Ontologies and
DBs).

∑ Demos: There were 22 submissions of which 19 were accepted.  The
demo chairs solicited other demos as well, but few of these solicited
proposals resulted in demos.  We decided on a demo plenary session
with 2 demos for presentation plus overviews of the rest by the demo
co-chairs.  The actual demos will be given (in parallel) during the
remainder of the session.

∑ Sponsorships: We contacted 48 organizations and 8 publishers.  Of
these, we received $25,000 from 10 sponsors/exhibitors, most at the
‘Bronze’ ($1,000) level.  We allowed Bronze sponsors free exhibit
space on a one-time basis to get more exhibits.  Not many publishers
(only ACM and MIT Press) wanted to exhibit this time.

Low Points/Suggestions:  

∑ Overall: It was not always clear whom to ask and who had the final
say when we wanted to innovate or to find out standard practice where
this was not specified in the ACL Conference Handbook.  It would be
good to clarify whether the NAACL Exec or the HLT conference board (or
the ACL exec) should be contacted for different matters, or to specify
which contact person in each the General Chair should deal with.  This
was particularly an issue wrt the Best Paper Award issue, the naming
of Area Chairs (the PC chairs wanted to call them Senior Program
Committee members), policies on co-located workshops (e.g. SigDial),
and many budget issues.  We got lots of help from lots of people, but
if experts and designated contact people could be specified in
advance, it would be helpful.

∑ Budget and fees: Over the years, responsibility for preparing the
conference budget and setting fees has been shared among the various
conference chairs and the treasurer of the ACL or chapter.  This needs
to be clarified in the ACL Conference handbook.

∑ Tax-exempt payments: A section should be added to the handbook about
the desirability of handling payments for the conference through some
tax-exempt organization.  Christy and I had assumed that ACL was tax
exempt and so originally were not budgeting for tax; we are now
hopefully getting Mitre to handle payments since they are tax-exempt
in Massachusetts.

∑ Registration: Holding the conference in early May may have lessened
the number of students who could attend.  However, there were problems
getting a good hotel later.  It is possible that this is a reason for
the low submissions to the Student Workshop.  However, despite
considerable efforts to advertise and encourage submissions, only 12
(9 NLP and 3 IR or Speech) papers were submitted, out of which 10 (7
NLP; 3 other) were accepted.  This is a problem.

∑ Website: ACL needs to move its website to a commercial hosting
service; this was a problem when we started registration.  In general,
the conference handbook needs to be clarified about all web issues wrt
conferences.  In addition, I would recommend that any workshop that
wants internet access should be able to get it, without paying for it
itself.  This is 21st century J.

∑ Co-located but non-ACL-run workshops: ACL and NAACL have developed
certain practices over the years in dealing with co-located workshops.
It would be very helpful to have the sections in the conference
handbook updated to reflect these and that future general chairs
follow them strictly.  Future conferences should offer to workshops
two choices: either the workshop conforms completely to ACL workshop
guidelines or it is completely on its own.  In addition, it would be
useful to make liaison with such conferences part of the Workshop
Chair’s duties.  There is too much duplication of effort otherwise.
This would require a change to the Handbook.

∑ Publications: Things went pretty smoothly despite several last
minute changes, one workshop getting its proceedings in very very very
late, and Omnipress mis-numbering the Companion volume (they reprinted
the TOC to match their numbering to fix this).  The Publications
Chairs handled all this extremely calmly.  If others have similar
problems, perhaps future conferences might want to find another press?
Also, the publications software we now use needs better documentation
in order to take advantage of some of its features.  This should be an
action item on perhaps for the NAACL exec.