JUNE 18, 2008

Date: June 18, 2008
Place: Hyatt Regency, Columbus, OH
NAACL Board members present:Jennifer Chu-Carroll, Hal Daume III, Bill Dolan, Graeme Hirst, Rebecca Hwa, Chris Manning, Owen Rambow, Suzanne Stevenson; Priscilla Rasmussen.
Guests: (For item 1) Kathy McKeown, Mari Ostendorf.
  1. President's 2008 report by Owen Rambow
  2. Treasurer's 2008 report by Christopher D. Manning
  1. Brief report on ACL 2008, and plans for NAACL/HLT 2009.
  2. Report from industrial relations subgroup.
  3. Site for NAACL 2010.
  4. Treasurer's report
  5. Planning expenditures for this year.
  6. Various conference issues.
[1] Brief report on ACL 2008, and plans for NAACL/HLT 2009.

Discussed the imbalance in coverage between the NLP, speech, and IR areas this year, but concluded that it's unrealistic to achieve a balanced split since speech and IR have separate communities and conferences. Noted the value of the high visibility invited talks across the areas.

In order to keep a reasonable submission and acceptance rate, need to add another (fifth) parallel session, another day of conference, or more posters; third option is most likely. One option is to have themed poster sessions. Discussed acceptance rate for posters: traditional for NAACL is 25% for long papers (which before 2008 have always been presented orally), 40% for short papers (which have always had a high percentage of poster presentations). This year the rates were equal, and there was some question about whether that is desirable. (Return to discussion of short papers later on agenda.)

Regarding organization for NAACL/HLT 2009: The General Chair will decide on non-PC chairs; the board doesn't need to vote on those. Noted the importance of having broad representation across the Americas. Some discussion of having a theme of the conference, having themed sessions, or just having Òpapers are especially encouraged in the following areasÓ.

Noted the need to clarify our (ACL/NAACL) multiple submissions policy, especially with regard to short papers, and also assure the visibility of the policy in all CFPs and submission instructions. (This year, a website error reduced its visibility.) COLING's policy is clear but we may want to ask them to clarify theirs to mention short papers.

(return to top of minutes)

[2] Report from industrial relations subgroup.

The industrial relations subgroup (Bill Dolan, Chris Manning, Owen Rambow) met with some industry contacts to discuss increased interaction with the NAACL community. In order to expand our research funding base (beyond US-government-supported projects) and to better serve our students (by further developing contacts with those who may provide future jobs), we need to more proactively establish ties with industry.

One idea discussed at length is the possibility of co-locating a future NAACL with the annual Text Analytics (TA) Summit, an industry conference with about 150-200 participants. One concern is that the latter is mostly attended by users and vendors, and doesn't have much technical content; the perceptions is that few of the people attending TA would attend NAACL sessions, so it wouldn't really achieve our goal of making industry more aware of our work and potential interactions. However, it might be a good opportunity for students to have exposure to the TA community. One option is to offer some tutorials that would be attractive to the TA community to encourage them to participate in NAACL. Other possible industry conferences could also be considered for co-location with NAACL: SpeechTek and SemTech; possibly others.

Other options for increasing NAACL-industry contacts (which we could do even in 2009) are to include in the NAACL program an industry panel, an industry track, and/or an industry demo session. A suggestion was to start with a focused tutorial, a focused workshop, or one panel within a session, and see what meets our goals. But the opinion was expressed that industry people aren't going to come for a single panel, track, etc Ð too much time and money; thus co-locating is the better option Ð that way, industry people would be on-site.

We need to poll more industry contacts on what they might find interesting and worth attending. For NAACL 2009, we could have 2 contiguous days of activities Ð eg, tutorial one day, panel or session the next. For 2010, we have to decide very soon if we want to try co-locating.

Action: The industry subgroup will continue to explore options for both 2009 and 2010, and circulate proposals in about 10 days.

(return to top of minutes)

[3] Site for NAACL 2010.

Lengthy discussion on the three bids for NAACL 2010, covering cost, geography (east/mid/west of North America), student housing options, appeal of location.

Discussion of broader issue: the way we do bids is awkward. We do a service to the community by allowing anyone to express interest in holding the conference, but we don't always get the geographical spread, appropriate hotels, etc.

(return to top of minutes)

[4] Treasurer's report.

Chris Manning summarized his written report [This will be a link to the file]. Our surplus is reasonably high (~$70K) and spending is relatively low (NAMCLO, JHU summer school). But keep in mind that our surplus could be wiped out in any particular year by a single conference having financial difficulty. ACL has to be in a financial situation to back the conferences if Òno one showed upÓ; that's $200-300K to cover a conference failure and start up the next one. But we don't have a firm guideline, so we must decide how much to spend each year. The treasurer's opinion is that we should have a larger surplus, and then determine how to distribute the surplus to benefit the field. We've been spending ~$20K/year. We could spend more, but to consistently do that we would need at least some conferences to have windfalls and that is not reliably predictable. We currently tell people to aim at a conference surplus of $10K, but we should increase that (the view being that that number was probably set when conferences cost much less).

Two small notes: We'll make $25/person extra on registration in 2009 because of the move to electronic journal format. We should put a time limit on reimbursements Ð give students 6 months to submit receipts for expensive; too difficult to keep track of over the long periods of time that some students take.

(return to top of minutes)

[5] Planning expenditures for this year.

a. JHU Summer School. $10K budgetted; 5 students applied, 5 approved; cost is $1000-$1500/person. We also agreed to fund taping of the sessions; approximately $1500. This keeps us within the $10K budget already allocated for this event. Note that the summer school needs to ok the taping with presenters; also, they shouldn't mike the students (to not discourage questions because they are being taped).

In future, should advertise the summer school earlier, to promote more students to apply.

For next year, we'll allocate $11.5K ($10K for student support, $1.5K for taping).

b. NAMCLO. It was noted that this is possibly good PR for our field and might attract women to CS. The organizers need some admin support, and the group needs more money for going to European competitions. We've been giving ~$2K; we agreed to contribute $5K this year (for student travel and admin support).

c. Funds to allow enable students to attend tutorials at NAACL. One suggestion was to give tutorial registration as part of conference volunteer package, but there were logistical problems with this. We agreed instead to establish a tutorial fund for students Ð stduents have to apply, and we fund as many as we can. We allocated $6K for 2009. Students will be chosen either first come first serve or lottery (TBD).

(We also discussed other funding for students to attend the conference [volunteering, Walker fund], and noted that we currently can't fund all the applicants for these sources.

(return to top of minutes)

[6] Various conference issues.

a. Short paper track.

We noted the strong sentiment from the ACL executive that the journal (Computational Linguistics) should be our publication of record, rather than the various *ACL conference publications. However, the ACL executive observed that very few long papers that are accepted by a *ACL conference are subsequently turned into articles for the journal, since many authors considered them as ÒterminalÓ publications, sufficient for the purposes of archiving and prestige. A short paper, on the other hand Ð especially one based on work in progress Ð will less likely be considered a terminal publication. The sequence of a short paper presenting work in progress, feedback at the conference (and early claim to the ideas), and a subsequent but fairly quick submission (as a short paper again) to the journal may be considered an attractive alternative to a terminal long paper at a *ACL conference, and may initiate a different way of thinking, on the part of researchers in our field, about the role of the journal and the conferences. Thus, part of the goal of short papers in the conferences is to encourage alternative types of papers that may later be appropriate for the journal.

The need was recognized by the NAACL board to clarify the goal of the short paper track and make clearer what we are looking for in the CFP. We would like to encourage novel things that can't appear in ACL/NAACL. The following categories were suggested:

1. `late breaking': what could be a long paper, presented in summary form; results not ready for the long paper deadline, which is earlier

2. truly short paper: completed work that needs only 4 pages for adequate description

3. not completed work (work in progress)

4. opinion pieces

5. negative results

6. `experience' reports, such as an interesting application

Concern was expressed about category (1): that this is simply a long paper that wasn't ready in time, and doesn't address the goal of encouraging different kinds of papers.

The coordinating of deadlines for short and long papers was discussed. It was noted that a staggered deadline (as used this year) allows a rejected long paper to be resubmitted, but it requires the long paper deadline to be earlier than would otherwise be needed. Given the emphasis on encouraging different kinds of papers to the short track (not just summarized long papers), this may not be an appropriate trade-off. Note that we may need separate reviewer pools if the two tracks are being reviewed in overlapping time, and reviewers need to apply different criteria.

We agreed on the following: Appropriate categories for short papers include those listed as categories 2-6 just above (i.e., not including Òlate breakingÓ). We will need to provide a better description in the CFP of what it means to be a short paper, reflecting the five categories we've identified. We should not delay short paper deadline to follow decisions on long papers: make short papers due 2 weeks after long papers (need 6 weeks reviewing time for long papers; 4 weeks for short papers). Separate reviewer pools should be established for the two, with distinct guidelines for reviewing the short papers (an appropriate review form or forms should be developed). We also need to clarify the multiple submission policy, since some unclarity and ambiguity were noted (more below).

We discussed whether to have submitters indicate for a short paper submission which of the five categories it falls in. This depends on whether we can develop a single review form to cover all 5 categories.

Action: Chris Manning will develop a draft short paper review form, aiming for a single form, and then we'll vote on whether submissions need to indicate which of the 5 categories they fall in.

We also discussed whether to make the names of short paper reviewers available to the CL journal to encourage short paper authors to develop the paper further for the journal. We agreed to do this, as long as only the journal editor sees the names.

b. Poster vs. oral presentation.

There is agreement that we should not equate long paper with oral presentation and short paper with poster presentation. We noted the issues with some Asian countries, where oral presentation is seen as much more prestigious, and recognized that we need to make clear that this is not the case.

We need to work with the NAACL 2009 organizers about the balance of poster and oral. Jennifer Chu-Carroll will communicate to them that we may want more posters (perhaps another session), and will inquire about what the options are. The goal is to encourage more papers, and the easiest way to do that is to have more posters.

c. Multiple submissions policy.

We agreed to forbid multiple submissions to the long and short paper sessions of NAACL: it would be contrary to the goal of encouraging different types of papers in short papers.

With regard to multiple submissions to the NAACL (either long or short) and another venue, we agreed that we need to clarify the current wording, especially with regard to short papers.

(return to top of minutes)