Conference on Computational Approaches to Linguistics?

A group of us have recently been discussing the possibility of a new conference on computational approaches to linguistics (group=Rajesh Bhatt, Brian Dillon, Gaja Jarosz, Giorgio Magri, Claire Moore-Cantwell, Joe Pater, Brian Smith, and Kristine Yu). We’ll provide some of the content of that discussion in a moment (we=Gaja and Joe), but the main question we’d like to get on the table is where the first meeting of that conference should be held. It’s so far agreed that it should be co-located with some other event to increase participation (at least for the first meeting), and the end of 2017 / beginning 2018 seems like the right time to do it. The ideas currently under discussion are:

  1. In conjunction with the Annual Meeting on Phonology in New York in early fall 2017. (We haven’t approached the organizers about this).
  2. In conjunction with a one-time workshop on computational modeling of language planned for fall of 2017 at UMass (invited speakers, pending funding, include Jacob Andreas, Emily Bender, Sam Bowman, Chris Dyer, Jason Eisner, Bob Frank, Matt Goldrick, Sharon Goldwater, and Paul Smolensky).
  3. As a “Sister Society” at the LSA general meeting 4-7 January in Salt Lake City (we have had preliminary discussions with the LSA and this seems very straightforward)

We’d very much appreciate your thoughts on the location or the substance of the conference as comments below, or use this google form to give a non-public response.

The original idea was to start a computational phonology conference, inspired by the success of the informal meetings that we’ve had as the North East Computational Phonology Circle, and by the central place that computational work has in phonology these days. But Giorgio pointed out that a broader meeting might well be of interest, and we seem to have come to a consensus that he’s likely right. It doesn’t seem like there is a general venue for computational linguistics of the non-engineering-focused kind, though we are aware of successful workshops that have been held at the ACL and elsewhere (e.g. Sigmorphon, MOL, CMCL). These workshops are in fact also part of the inspiration for this; however, the conference we envision would be broader in scope and co-located with a major linguistics conference to attract as many linguists as possible, minimize costs, and minimize additional conference travel.

We still think that a core contingent might well be the computational phonologists, especially at first, so we still think co-locating it with AMP might make sense (plus NYC is a good location). We’ve also had suggestions that we might in some years co-locate with other conferences, like NELS – the location of future meetings is something we could discuss in person at the first one.

We also seem to have come to a current consensus that we’d like to have reviewed short papers in the CS / CogSci tradition. This is an extremely efficient way to get research out. The one worry that was expressed was that this may create a barrier to later journal publication, but at least two journals with explicit policies on this (Cognitive Science and Phonology) allow publication of elaborated versions of earlier published conference papers.

Please share this post or the tweet below!

88 thoughts on “Conference on Computational Approaches to Linguistics?

  1. Robert Daland

    I have a thought and a clarification request.

    THOUGHT: The folks at ACL know that Computational Linguistics has drifted away from what generative linguists are interested in, and a percentage of them want to keep the bridge open. It might be nice to involve them in this discussion, from a couple of angles. The first is that we might benefit from their perspective — how their efforts toward this have or have not worked out in the past and why, who to reach out to maximize awareness and participation from the ACL community. The secondis that reaching out might avoid duplication of effort, or worse yet dilution. A result that we do not want — sorry, that *I* do not want — is that generative linguists start staying home from SIGPHON and ACL-Cognitive Computational Modeling and other ACL-adjoined workshops. One name that springs to mind is Michael White. I’ve forwarded this blog post to him and to Tim Hunter, UCLA’s newest faculty member.

    CLARIFICATION: “consensus that we’d like to have reviewed short papers in the CS / CogSci tradition”. Does this mean that when people submit, they submit papers instead of abstracts? Or that they write papers later, and possibly not all of the abstracts that were submitted become papers?

    Reply
    1. Joe Pater Post author

      We were thinking submitted papers. And thanks for spreading the word to broaden the discussion (I don’t know Michael, but did also send to Tim).

      Reply
      1. Thomas

        The ACL has made efforts in the past to stop the “engineer-y-fication”. They have not worked out well, though I don’t know any details. Jim Rogers would be the person to ask, he was on one of the committees. Basically, even MOL, which is an ACL SIG, is the niche of a niche at the ACL as it is right now. However, the Transactions of the ACL (TACL) seem to be more diverse, so there’s hope.

        Reply
  2. Thomas Graf

    For venues, I would generally urge to keep it in North America. It reads like this is the idea anyways, so the next few lines are just to reinforce your stance 😉 There’s already tons of computational linguistics conferences outside the US: FG is co-located with ESSLLI, LACL is always held in France; MOL and tag+ mostly alternate between Europe and the US, and the ACL is all over the world. Going through my list of presentations over the last 8 years, almost all computational work was presented in Europe. For me that wasn’t too much of an issue since I go to Europe every summer anyways, but that is not an option for most people. So by keeping it a North American conference you would help the field as a whole, and it would set you apart from the
    “competition”.

    One thing you might consider is co-locating the conference with the LSA summer institute or NASSLLI (depending on how soon you want to get the conference off the ground, maybe even for the very first meeting). In contrast to ESSLLI, NASSLLI is only held every other year (next 2018), so this would still allow you to co-locate with linguistic conferences on a regular basis. Summer schools draw much more students than any linguistics conference, and attracting new students from all over the country to compling is at least as important as promoting our work to established theoretical linguists.

    I would also include *Glossa* and the *Journal of Language Modelling* among the choices for journal publications. The latter is definitely fine with that (the revised MOL proceedings will appear there as a special issue), and I would imagine that Glossa is open to it, too. Both are open access and thus deserve special consideration.

    One final comment on going beyond computational phonology: if done correctly, this would make the conference stand out in the field. FG, MOL, LACL, and so on are on the technical side (I wouldn’t say engineering, but there is a higher than usual expectation of mathematical rigor). It would not feel right for me to submit a paper to those conferences that is formally rather trivial but addresses a point of linguistic interest. For example, I once gave a talk at LACL showing that Late Merge is just a restricted form of lowering movement — a syntactician would find that troublesome, but with the LACL audience it fell flat because they don’t care that much about Late Merge or Lowering to begin with. Similarly, the MIT workshop on MG parsing last October had several interesting talks that I don’t think would’ve made the cut at the more formally minded conferences. So something to fill the gap between a normal linguistics conference and the venues for mathematical linguistics would be nice.

    However, there is also a big risk there. I believe computational
    phonology (and to a slightly lesser extent computational morphology) has a lot of cohesion regarding the formal models. If you know OT, some statistics and the main topics of an intro class to computational linguistics (n-gram models, finite-state automata, HMM) you can roughly understand what other people are doing, though the details may be opaque. I’m not a phonologist, and yet I’ve never had a problem understanding a computational phonology talk at least in broad strokes. Computational
    syntax is much less uniform. You have very different approaches (MGs, dozens of variants of TAG and CCG, Lambek grammar, Dependency grammar, …), none of which is more dominant than the other. You need tons of formal background that isn’t necessarily part of the normal compling curriculum (tree transducers, weighted tree automata, various logics, algebra, parsing algorithms, …). I’ve been a computational syntactician for 8 years now, and I still hear talks once in a while where I don’t understand a word, or why the results matter. Since the conference won’t be big enough to have parallel sessions, this diversity really poses a threat to its cohesiveness. So if you decide to go beyond phon & morph, it will be important to pitch the CFP such that you get submissions with wide cross-audience appeal. Maybe call it something like CompPhon+, which allows for submissions from other areas but also conveys clearly
    what the average audience member looks like.

    Anyways, I think this is a great idea that could have a very positive effect on the field.

    Reply
  3. Ewan Dunbar

    Contra Thomas, I would urge you to not keep it inside this narrowly North American theatre. Keeping it a North American conference serves no useful purpose, especially if you’re proposing to make it a conference that’s actually worth a paper.

    Can I suggest a strong none of the above? The LSA is a non-conference for those outside the US (it only just registers in Canada and even in the US a lot of people make a point of not going); AMP is only attended by phonologists; and option 2 sounds like an excellent option for the first year, but that’s passing the buck on what to do next year.

    Think bigger? Co-locate it with ACL or CogSci. The advantage of ACL would be that we could eventually squeeze out and absorb SIGMORPHON (maybe even MOL) and build a bigger conference. SIGMORPHON is really too small to exist at this point. The advantage of CogSci would be that the registration fees are much lower. And the advantages of both are obvious – they rotate around the world rather than just around places within spitting distance of the east coast; a lot of people will be going to one or the other anyway; and they’re much bigger and higher profile conferences than any of the ones you mentioned, so you’ll attract more people a priori.

    Reply
    1. Ewan Dunbar

      As an afterthought based on a couple of suggestions here, I would add that, on the theme of having as a target “relatively high-profile, important, relatively large, international conference” – specifically the international part – I think (but don’t know) the LSA *institute* is attended by a lot more internationals than the LSA *conference*. So that might be another reasonable option.

      Reply
      1. Thomas Graf

        Just wanted to point out that I explicitly talked about the LSA summer school. I fully agree with you that the LSA conference is not a good venue.

        And my recommendation for keeping it in NA is due to the fact that there are already several good conferences that are always or at least regularly held in Europe (and NA and Europe are where you find the majority of computational linguists): FG, LACL, MOL, tag+. Computational linguistics conferences are already more expensive than those in theoretical linguistics (we’re talking $250 vs $50 for students). I would love to send my students to FG and LACL each year, but that’s over $2,000 per person. If MOL and tag+ don’t happen to be in the US a given year, there’s no affordable places for my students to present their work. You might say NAACL, but that’s just as engineering-heavy as the ACL. I’m not saying that these issues must be taken into account when picking a venue, but they are certainly pertinent enough to merit discussion.

        I also agree with Greg on MOL, FG, LACL, and tag+ not being engineering focused, I didn’t emphasize that enough in my first post (in fact, I barely mentioned it at all). They are open to linguistically minded work as long as it has a formal grounding, and a lot of the work that currently runs under the banner of computational phonology could be presented there. What I wanted to highlight with my example, Greg put much more lucidly: some topics are a hard sell because implicit assumptions are not shared by all audience members.

        Reply
        1. Ewan Dunbar

          Yeah, sorry – on the LSA conference point I had only Joe’s initial comment in mind, I was meaning to just +1 what both you and Greg said about the institute. I also didn’t mean to say that not saturating the European market wasn’t a legit reason. I meant more that setting up a conference in North America by design (or equally in Europe by design) doesn’t seem to have much of a point. And, well, it’s annoying. The only potential reason is travel cost/time, and co-locating it with a large conference people are already likely to go to seems to me to be a good way of mitigating that cost. Another thing that will help in the positive direction will be that it delivers a real peer-reviewed conference paper. Both of those things together, plus a strong start that makes me think that a lot of good people will be there, and I’m likely to go. But co-locating it with a large *already international* conference that I’m likely to go to seems to just add points in the right direction, so why not?

          Co-locating with an existing huge conference (SfN) worked so well for Neurobiology of Language that they spun off within a few years. SfN is US-based, but it’s a US-based conference that literally every neuroscientist goes to. We don’t have an equivalent. In fact, we don’t have an *international* conference that literally every technically-minded person interested in language goes to, or is primed to go to. But whatever comes closest to that would be my target. I know not a huge number of linguists go to Cog Sci, but among the math/tech subset of linguists I suspect the proportion who do is much higher than among the general population. CogSci also presumably has higher attendance overall than the LSA institute due to its nature as a proper conference.

          It’s amazing how much SIGMORPHON doesn’t attract any casual visitors and submissions from the rest of the ACL, keeping the communities at arm’s length. I suspect that would be less the case at the LSA institute, and at Cog Sci, pulling in some different folks at each.

          Reply
  4. Greg Kobele

    MOL, FG, LACL, and TAG+ are decidedly **not** engineering oriented — they typically present results showing that certain configurations of linguistic assumptions derive systems with certain properties, as opposed to presenting simulations where from certain initial states one can arrive at a state which is good in some way. I think that the results presented at these conferences are in fact of direct importance to linguistics; there are however implicit assumptions made on both sides (what counts as notation vs substance, what assumptions are linguistically relevant) that obfuscate the fact that our two communities share goals and vision.

    There is great interest among the regular participants of these conferences to draw connections to and make alliances with linguistics; for example, [last year’s MOL](https://lucian.uchicago.edu/blogs/mol2015) was colocated with the LSA Summer School in Chicago.

    I agree with Thomas, however, that a venue for bringing out the linguistic relevance of MOL-like results would be useful in bridging this perceived gap. I suspect that there are more P-side people with interests in formalization than S-side people, and thus presumably this would be reflected in the distribution of acceptances arising from an open call for papers. I’d worry, therefore, that emphasizing this in the name (CompPhon+) might exacerbate this already existing difference.

    Agreeing partially with Ewan, I think that option 2 (colocating it with the UMass workshop in 2017) is the right solution for the moment; finding a principled resolution to the broader question can certainly wait.

    Reply
  5. Joe Pater Post author

    Thanks everyone for your thoughts about this!

    One of our main goals is to make this a conference that particularly accessible for linguists. It’d be great to have a conference that graduate students with just a burgeoning (or passing) interest in computational work would go to, get exposed to the cool research, and grow their interest. Co-locating with the ACL wouldn’t work for that, because it’s so expensive, and generally inaccessible in terms of content for most linguists. I’m guessing that there’s room for at least two conferences of the type we’re talking about – one that leans computational / mathematical and is held with the ACL, and one that leans more linguistic. And maybe one that leans more towards cognitive modeling (processing, neural) and is held with CogSci. It seems like there are existing workshops that are in the general ballpark that we’re discussing at ACL and CogSci – though maybe amalgamation or generalization at ACL would be a good endeavour. (Just to be clear – we were citing SIGMORPHON et al. as non-engineering focused initiatives that we want to build on – Greg’s comments made me worry we might have been misread). Maybe in the future we’d want to go to ACL or CogSci to spread the gospel and expand in those directions, but it doesn’t seem like the right move for our first meeting at least.

    The Summer Institute is a good idea in principle, but in 2017 it’s in Kentucky, and the deadline for workshop proposals is Sept. 1. Both of these count against it for me, anyway.

    Our google form has elicited 4 responses, from 3 syntax/semantics people and one phonologist (all people I was very happy to hear from). 4/4 would go to UMass, 3 to NYU and only 1 to the LSA in Salt Lake City. I was leaning heavily towards the LSA option before this discussion, but now am ready to drop it unless I hear objections.

    I’m currently leaning toward UMass, given the interest shown in the poll and in the comments. We’ve got a great lineup of speakers at the workshop to build on, which probably explains the interest people have (New York would likely be the winner otherwise, unless we’ve got a lot of leaf peepers out there).

    Reply
  6. giorgio

    We seem to all agree that it is a good idea to co-locate the first edition or the first two editions of this new computational linguistics conference with some big linguistics event(s). This is to achieve two goals. First, to make it explicit that this is a *linguistics* conference, just like other linguistics conferences, like SALT or NELS. Second, to make the new conference more visible, to spread the voice, to attract people, to build critical mass. Which of the co-locating options considered so far would help us to best achieve these two goals?

    I am skeptical of the idea of co-locating our new computational linguistics conference with the Amherst workshop. Of course, I do find the topic of the Amherst workshop really interesting and exciting. Yet, it is an occasional and thematic workshop. I am not sure that co-locating the new conference with such an event would meet the goal of boosting the visibility of the new conference. I think we need to co-locate the new conference with some big linguistics event that all or at least many linguists track on their calendars.

    I am skeptical of the idea of co-locating our new computational linguistics conference with the ACL (trying to replace Sigmorphon) or the CogSci annual meeting. That would defeat the purpose of establishing this new conference as a linguistics conference on equal footing with, say, SALT or NELS. Who would co-locate, say, SALT with CogSci? (This does not exclude, of course, that one of the invited speakers could belong to the ACL’s crowd).

    I am skeptical of the idea of co-locating our new computational linguistics conference with the LSA meeting in January. I think the LSA meeting is a crazy place, it hosts events of very mixed quality, and many respectful linguists tend to snob it (and brag they have never attended it).

    In conclusion, I like very much the idea of co-locating our new computational linguistics conference with some big linguistics conference such as AMP or NELS. Actually, my preference would go for NELS (does anyone know where NELS 2017 will be?), because that would help us spread the message that, although lots of computational work in linguistics happens nowadays on the P-side, we do want to construe the scope of the new conference more broadly than the P-side. Otherwise, the idea of co-locating this new conference with AMP seems rather good to me and I am confident that people at NYU would be supportive.

    Reply
  7. Joe Pater Post author

    One issue with NELS is that it’s not clear that it’s a conference that a significant portion of the group we are targeting as participants goes to regularly. Another is that it’s a purely student-organized conference, which makes the process of planning the co-location more difficult. So I’d prefer AMP over NELS, given that at least a lot of computational phonologists are going there.

    As well as the goals Giorgio outlined, another goal for the first meeting is community building, getting the people together who will constitute the core participants in the future. I’m a little worried that doing it with AMP will tilt it too much towards phonology, when we’re aiming for something broader. That said, New York is an accessible place, so maybe people will come anyway. Of course, if we co-locate with AMP or any other conference (besides the LSA), we’ll likely have to schedule it right before or right after, and this will not be the best for getting people to come.

    Given that last point, we should also think of doing this from the start as its own conference on its own weekend.

    Reply
  8. Ewan Dunbar

    The ideal conference that we would want to co-locate this with in order to get a reasonably large set of submissions of quality broadly-computational papers, plus attendance by linguists, doesn’t exist and is perhaps exactly the conference under discussion.

    I can think of two possible reasons to co-locate it. One is that if it’s far and it isn’t an established conference then you’ll lose some of your core attendees in the first couple of years, which could be a bad start. The second is if you want to attract linguists you might get some of them wandering over if they were already at another conference.

    Both of these conjectures to me need a bit more careful scrutiny.

    If the call is well written (expected to attract good papers) and it generates a publication, maybe the first worry is less important.

    To attract linguists, wouldn’t it be best to invite linguists? If there’s no room for the working syntactician in the call, they’re not likely to come over to the conference, either.

    The only place I can think of is that the LSA institute is usually in a pleasant location, is full of intellectually keen, open, and international linguists. Including students, the key victims. It’s bigger than AMP, NELS, or SALT, and isn’t restricted to one subfield. It’d require some scheduling coordination, I guess.

    Reply
    1. Joe Pater Post author

      I agree that the LSA summer institute is an excellent venue (though it’s bi-annual). Submitting a workshop proposal for Sept. 1 for the 2017 edition in Kentucky is still an option – just one that I’m not enthusiastic about for personal reasons. There isn’t a “Sister Society” co-location option like there is for the winter meeting.

      And what about Europe in the summer? I’d personally prefer that, and there might be a significant number of others. We could rotate it between the two perhaps (this picks up the thread about location continent-wise – our original suggestions were just those that came to mind, I believe).

      Reply
  9. Thomas Graf

    One thing to consider is that the way the proceedings are published also signals a certain alignment to interested parties. Linguistics doesn’t have much of a tradition of peer-reviewed conference proceedings, whereas computational linguistics does. So one could pick the venue with linguists in mind and the proceedings outlet with computational linguists in mind so as to attract attention from both communities.

    For example, publication in the ACL anthology (that’s possible without being co-located, see e.g. tag+) with a follow-up issue of heavily revised and extended papers in, say, the Journal of Language Modeling, would look very attractive to computational linguists. Add some well-known invited speakers from both communities, compile a diverse board of reviewers (and make their names public, as is customary for MOL, FG, …), advertise the event widely (LinguistList, FOLLI and MOL mailing list, ACL wiki, etc), and you should be able to peak the interest of many researchers.

    Reply
    1. Ewan Dunbar

      I +1 all this. I also put forward explicitly what might have been implicit before, which is that the publishing model of submitting a whole (for linguistics, short) paper in advance, and having it peer reviewed before conference acceptance, is, despite being unfamiliar to linguists, worth proactively introducing.

      Reply
  10. Naomi Feldman

    I guess my main question is what range of work you’d be targeting with this conference. In addition to work that’s solidly on the linguistics side, are you also hoping to attract work in computational psycholinguistics (which I see as being situated between cognitive science, computer science, and linguistics)?

    My students and I often present our work to linguistics audiences, but so far, we’ve always also been able to present the same work again at CogSci or ACL. It seems like what you’re describing here would force us to choose between these audiences, since (although I agree it’s OK to write a subsequent journal article) it’s taboo to write two conference papers on the same work. This seems like a big drawback to me — computational psycholinguists are already spread very thin, and this would spread us even thinner. Is there any way to increase our visibility to linguists without reducing our visibility to cognitive scientists and NLPers?

    Co-locating with ACL or CogSci would be great from my perspective, though I understand the concern that it wouldn’t reach as many linguists. But wouldn’t it be nice if linguistics became a major force in the CogSci community?

    (It’s also possible that the need to reach out to NLP and computational cognitive science audiences is greater for those whose gender features make them less representative of those communities; so this may be a bigger issue for me than for others.)

    Reply
        1. Joe Pater Post author

          Not sure what the co-locating options are – didn’t see anything obvious on the website. And of course, there have been some reservations expressed about CogSci as the right venue for this endeavour. But in any case, I’ll be trying to get to London next summer!

          Reply
          1. Naomi Feldman

            There are a couple options. One, it could be a CogSci workshop, submitted around the same submission deadline as papers and then (if accepted) with a separate call later to get submissions to the workshop. But then there are also a bunch of conferences that aren’t CogSci workshops, but are co-located, e.g., see the ones listed at the bottom of the page here: http://cognitivesciencesociety.org/conference2016/index.html

            MathPsych (http://www.mathpsych.org/conferences/2016/) is a good example of a conference that typically co-locates with CogSci.

    1. Claire Moore-Cantwell

      On the topic of trying to get more linguists to be a part of the CogSci/ACL communities: It strikes me that this isn’t going to be as easy as co-locating with those conferences, especially if the submission format is something linguists aren’t used to. If this should be one of the goals of the conference, perhaps it would make sense to specifically co-locate with both kinds of conferences? Perhaps start out co-located with a major linguistics conference (NELS/AMP/LSA/LSA institute), and then move to CogSci or ACL the next year? The conference could even bounce back and forth between the two communities. The downside of that would be missing this quite relevant upcoming CogSci. If linguists attend, and like, the conference when it’s at NELS/AMP/LSA, then perhaps they’d consider attending it when it’s at CogSci/ACL even when they otherwise wouldn’t.

      Reply
      1. Thomas Graf

        I’m not sure that the format itself is something linguists aren’t used to. Several linguistics conferences have proceedings, e.g. CLS and NELS. So linguists know how to write a 12 page paper, they’re just not used to writing it before the conference. I don’t think that’s a major issue, but maybe Robert’s clarification question was prompted by a strong dislike for this model.

        This raises another issue that limits the choice of conferences: if submissions must be papers rather than abstracts, then you have to make sure that the paper deadline isn’t in the spring. All the computational conferences that have been mentioned so far (ACL, FG, MOL, tag+, Sigmorphon, LACL, maybe CMCL) usually have their deadlines in March and April, so this is already a very stressful period.

        For a conference that takes place in June or July, you can’t have the deadline much later than that, and an earlier date would also seem odd (“Why do you need the papers so far in advance? I’ll barely remember what I wrote by the time the conference comes around.”).

        Reply
        1. Ewan Dunbar

          (Sorry Claire, I missed this thread before when commenting a +1 for the format, in spite of its unfamiliarity, above.)

          Reply
  11. Gaja Jarosz

    Thanks so much everyone for all your comments – this is really useful information and a lot to think about. The primary goal of this meeting as I see it is to build on the success of NECPhon, establishing a quality, refereed venue for linguists (and psycholinguists) who do formal or computational work. We want this to be as accessible and attractive to linguists as possible, including students. This means affordable and with enough linguistics content (possibly due to a co-located linguistics event or due to sufficient length of the event itself) for linguists to come. For me, co-locating with ACL is not an option for these very reasons – despite their inherent interest, workshops like sigmorphon have not been enough on their own to attract linguists. The conference is too expensive, the material at ACL is not accessible to most linguists, and there is not enough linguistics content at the event overall to warrant the expense or the time for most linguists to go. Obviously, that creates a cycle where linguists don’t submit their work there because they know no linguists will be there to hear about it.

    Before we solicited this input, I was leaning toward co-locating with the LSA. We need to build a critical mass of linguists and encourage linguists who might not specifically seek out a computational conference to attend. At the same time, we obviously want the conference to attract those of us who are already doing work in the area. There are only so many conferences one can go to each year, so co-locating with an event that many linguists go to anyway makes it all the more likely to build the critical mass we need to make this successful. However, what I’m hearing from a number of you is that the LSA is not appealing for various reasons, so I’m ready at this point to cross off that possibility too.

    While I would love to see linguistics become a major force in Cogsci more generally, my suspicion is that a 1 or 1.5 day workshop at Cogsci would not attract very many linguists. My impression is that only a small percentage of computational/mathematical linguists attend cogsci regularly, and certainly there are few linguists there. The year I went there wasn’t a single linguistics talk and very few about language. I’d be happy to be proven wrong on this, but based on my current impression, I would only consider possibly co-locating with CogSci once we already have a critical mass; it doesn’t seem like a good choice for establishing this as a major event for linguists.

    I would like this to be an event that is attended by international scientists, so for that and other reasons Joe gave, I think NELS is not ideal either. So that still leaves AMP, UMass, and LSA Summer Institute on the table for me for the first meeting. I also like the idea of alternating US – Europe every other year, making the conference accessible to everyone at least every other year while being inclusive of the international community.

    AMP would be perfect if this were just a computational phonology conference. I would be really grateful to hear from everyone on this issue: should this be just a computational phonology conference? Reading in to the comments above, I see suggestions going both ways. On the one hand, I think many of us are quite confident that this will be successful in computational phonology, and that we could have an interesting and cohesive conference in this subfield. It’s less clear how much interest there would be in a general computational linguistics event and whether making the scope that broad could actually dilute the attractiveness of it for some people? On the other hand, it sure seems like we could use a general computational linguistics conference for linguists.

    On the UMass option, I can actually see an advantage to the fact that it’s a one-time event. That means we are not associating ourselves in any permanent way with a particular meeting (or giving that impression). And the line-up of invited speakers is so good and is also in computational linguistics that I am hopeful we could get a lot of interest from exactly the core community we want to target. Joining forces with the UMass event would also make the whole thing a 3(or so)-day event in computational linguistics, which sounds pretty attractive to me at least. None of the other options would have that much computational linguistics. We would then need to immediately decide what the future meetings should look like, and if the right core community is there at the first meeting, that would be a good place to have that discussion.

    I do also like the summer institute option. I have never been, but I’ve been wanting to go, and I do get the impression that people who have been have found it to be an intellectually stimulating experience. The summer time is nice, but if we alternate with Europe every other year, where could it be summer 2018? ESSLLI?

    Reply
    1. Naomi Feldman

      It’s true that only a small percentage of computational/mathematical linguists attend CogSci regularly, besides computational psycholinguists. On the computational psycholinguistics end, though, it’s the opposite — I’ve never been to NELS or LSA (conference or institute). I’ve been to AMP a couple times, but never submitted there.

      If pitched right, I agree with you that co-locating with the UMass event could draw in many of the relevant communities.

      Reply
    2. Ewan Dunbar

      I love the idea of alternating US-Europe. Particularly since I have the impression that the LSA institute pulls in a non trivial number of Europeans already. While I’d be happy to make a trip to the US every summer around the institute, I’d be even happier not to have to. ESSLLI in the abstract sounds like a good place to try and put it, but I’ve never been.

      Reply
      1. Thomas Graf

        I have a very pre-mature nitpick: If alternation, I would like to see a slightly more complicated alternation pattern to ensure complementary distribution with MOL. That way, both Europe and the US are well-served.

        What makes things complicated is that MOL only takes place every other year. So MOL is on the same continent every 4 years. Strictly speaking there isn’t even an official alternating schedule, and MOL has been in Japan before. But the US/Europe alternation seems to be the intended pattern. So if we do not want to overload any given continent with computational conferences every four years, we’ll need some kind of leap-year rule.

        But as a side, that’s a very minor thing that’s not really worth worrying about at this point. Just wanted to point it out in case Europe/US alternation becomes the plan for the future.

        Reply
      2. Thomas Graf

        Regarding co-locating with ESSLLI: this has the problem that ESSLLI is already co-located with FG, and since both are run by FOLLI, I’m pretty sure that FG would always get the prime weekend spot between week 1 and week 2. So you’d be left with the weekend before week 1, which doesn’t draw a big crowd — I believe FG had that spot until 2007, and then they decided to change it because only FG-participants would show up.

        To be honest, I’m not sure that it has helped much, and the reason is rather simple: ESSLLI is already a tour de force for students. There’s tons of courses, evening lectures, and social events, so it takes real dedication to spend your weekend at a conference instead of a sight-seeing tour with new friends or simply in your bed.

        Reply
  12. Joe Pater Post author

    I see Albright, Goldsmith and Magri are all teaching courses next summer at the LSA:

    http://lsa2017.uky.edu/courses/

    Workshops are supposed to tie into courses, so that would be easy to do (plus we’d have a core of attendees).

    So this seems doable, but let me give one more argument for the UMass option. We’ve submitted an NSF grant application to cover our speakers, and we’ll be going after industry and university funding for other expenses. Presumably, we could tie at least some of the expenses of this conference into the funding requests we make in the future. We’d be starting from scratch in trying to fund the LSA workshop – and it’s now later than the NSF likes for proposals.

    Reply
  13. Giorgio

    I think that co-locating the new envisioned computational conference with the UMass workshop sends the wrong message about proportions. I see this new event as a future big, respected, regular, international conference on a par with AMP, SALT or NELS. One should not co-locate a big regular conference with an occasional thematic workshop (no matter how excellent the workshop is). The proportions are just not right. If any co-location should happen, it should be reversed: the UMass occasional thematic workshop should be co-located with the new big, regular conference.

    I definitely see Joe’s point that co-locating the new envisioned conference with AMP might not be a good idea because it would make the new conference too phonology-focused while co-locating it with NELS might not be a good idea because it might not help us attract sufficiently many phonologists! It looks like we simply cannot find the right conference to co-locate with.

    So, I would like to revise my initial opinion and submit that perhaps we should not co-locate the new conference with any other conference or workshop. Perhaps, we should be daring and bold and start to walk with our own legs right from the beginning. This would allow us to choose freely and autonomously the period of the year that we think would be most propitious (the fall? the winter break? the spring? the summer?). And it would allow the new conference to be scheduled on a Friday+Saturday (perhaps a little bit of Sunday) as all respected conferences.

    Let me add that (for obvious practical reasons as well as for deeper sociological reasons) I like very much the idea of the new conference alternating between the US and Europe. And I would like to formally bid for organizing the second meeting of the conference in 2018 in Paris–I do think that the first meeting should be in the States where the critical mass is larger.

    Reply
    1. Joe Pater Post author

      I disagree with Giorgio about the optics, but am not at all opposed to the conference being held independently, if there is a host willing to step up for 2017, and if people are interested in going to another independent conference every year. I should also say that when we set up AMP, we made sure that there was about a half dozen potential hosts. For us, that meant a reasonably sized contingent of graduate students who could be counted on to help in the organizing. In phonology, we were able to count enough programs that could be potential hosts by that metric. Is the same true in computational linguistics, of whatever kind we’re talking about here?

      Reply
      1. Naomi Feldman

        I’m still concerned about being spread too thin, in terms of having too many different audiences that aren’t mutually compatible. You can only publish one conference paper and one journal article — and need to optimize those for the relevant broad audiences as much as possible. It’s not clear that a paper from a new linguistics conference should or can win out over CogSci or ACL workshops (let alone ACL main conference).

        The benefit of co-locating with the UMass workshop (or CogSci or ACL) is that it could attract the relevant portions of the audience from those other communities, and make it less risky to forego their already-prominent, well-attended yearly conferences.

        Reply
        1. Joe Pater Post author

          This fits with the “neutrality” of the UMass option that Gaja pointed out. We’re not committing right away to being an independent conference, or to co-locating with any particular existing one. It seems that people have a range of views on these issues, and spending some time hashing that out, with full group of people who show up at the meeting, seems like a good approach.

          Reply
        2. Gaja Jarosz

          I’m not really sure I am following what you are suggesting, Naomi. I don’t think it makes sense for us to try to compete with CogSci or ACL – I think all we can hope for is that people who go to these conferences will also sometimes go to this one. Isn’t the fact that there are three different audiences just a reality for anyone who works at the intersection of these three fields? If you already have to go to linguistics conferences to reach linguists (since they are not at CogSci or ACL), wouldn’t this event be a good choice for reaching computationally oriented linguists? Also, I think there’s a large contingent of linguists who don’t go to ACL or CogSci but who would potentially go to this conference.

          Reply
          1. Naomi Feldman

            If you want people like me to be an occasional attendee (rather than a regular one), then I agree it doesn’t matter. That’s why I initially asked who your target audience was.

            But note that so far I’ve dealt with the “three fields” problem by being in a linguistics department, but publishing in cognitive science and NLP venues. NECPhon is no problem under this model, but the new conference would be, I think.

          2. Gaja Jarosz

            Of course, I’d like you to be a regular attendee! But I’d also like linguists who don’t go to ACL or CogSci to be regular attendees. Is that possible? Are you saying that this conference would be more practical for you if it had only talks and no papers/proceedings?

          3. Joe Pater Post author

            It seems there would be ways of finessing the proceedings issue you raised earlier Naomi. For example, in conferences with post hoc papers (e.g. NELS, AMP), one always has the option of not publishing in one of the proceedings, but presenting at both. Perhaps we could entertain a similar model. Or even more creatively, one might imagine trying to negotiate the right to publish the same paper in multiple venues – I don’t see a real issue with this if the audiences are really largely distinct (so long as one is not trying to get multiple credits for the same work). With electronic publishing, there would seem to be ways to make this feasible.

          4. Naomi Feldman

            For me, yes, having optional proceedings would make things simpler. But for others, I’m not sure — for example, I’m not sure if something like that can be part of the ACL anthology.

            My favorite example of two conference papers on the same work is this one:
            http://homepages.inf.ed.ac.uk/sgwater/papers/acl06.pdf
            http://homepages.inf.ed.ac.uk/sgwater/papers/bucld07.pdf
            but I think the audiences we’re talking about would be similar enough that it’s a bit sketchy to try and publish at both venues.

          5. Thomas Graf

            The problem with optional proceedings is that initial submissions would have to be abstracts, no? I’d be rather skeptical if a computational conference admits based on abstracts, simply because this is not a good format for evaluating most computational work (nor most linguistic work, but the field apparently disagrees). Or is the idea that everybody submits a paper but can opt not to have it published in the official proceedings? I’m not sure how I’d feel about that, it’s definitely unusual.

            I’d much rather see the authors publish multiple papers at different venues that focus on slightly different aspects of the work and are finely attuned to the background of the respective audience. Just like the example Naomi gives. I don’t think that this is somehow impossible for this conference due to a large audience overlap or something like that. If you look at Jeff Heinz’s work, for example, there is a clear difference between the papers written for computational linguists and those written for phonologists, even technically inclined ones. So it all comes back to defining the intended audience and the mission statement of the conference, which seems to be the big question mark at this point that decides all the other issues (proceedings, co-location VS stand-alone, venues).

          6. Naomi Feldman

            I seem to recall seeing at some CS conference that there was an option to submit a paper that had already been published at a (substantially) different conference. It would then be reviewed and if accepted, the authors would present the same work at this conference, but wouldn’t publish in the proceedings. I’m trying to remember where I saw this, but now I can’t find it.

            Maybe something like that could be an option here?

            Other venues have clear guidelines that conference papers need to be original, unpublished work. Even in the example above, the ACL paper was published before the BUCLD one.

          7. Joe Pater Post author

            @ Thomas: I think the current consensus is a conference aimed at a linguistics leaning audience, with submitted papers, perhaps appearing in the ACL anthology. This seems like a great way to achieve cross-fertilization across disciplines, while establishing this conference as something new with a distinct focus that fills what seems to be a real need. The notion of optional publication is unusual in this format, but I don’t see a problem, especially if we could simply link in our proceedings to the paper published in the main ACL session or wherever. What this would mean is that we would be willing to accept simultaneous submission with other conferences – the current practice amongst linguists (at least some), but not allowed by the machine learning conference Naomi links to. The general danger of this set up is that people would submit to multiple conferences and only go to one, and waste reviewing resources, but I don’t think that’s likely for this.

  14. Colin Phillips

    Sorry to be late chiming in (thanks to Naomi for alerting me to the discussion). I’d like to put in a word for co-locating with the LSA January meeting, recognizing that it is not a popular option among those who have weighed in so far. Some reasons:
    — Co-locating with a big meeting is a plus, as many agree
    — For all its past ills, which are well known, LSA is the only conference that has a realistic chance of being a venue that brings together the whole field
    — LSA is professionally run. No need to worry about accommodation, venues, etc.
    — LSA has been making a real effort to meet in more attractive venues, while holding to its long-standing commitment to keep costs low for students. (Venues are reserved 5 years ahead of time, so the Minneapolis error was planned loooong ago.)
    — LSA provides other services that other meetings don’t provide, e.g., increasing professional development activities for students
    — LSA would likely be motivated to cooperate on this, seeing the value for the field (and for the LSA).

    Yes, yes, I know that LSA has a bad reputation. Mostly from people who haven’t been recently. If anybody tells you that it’s all about the job market, they probably haven’t been there since Skype was invented. Yes, it is in early January, but that’s how you get to offer conferences in nice venues where students can get a quad room at a low rate with free wifi thrown in. Even 2 weeks later an the prices would skyrocket. LSA is still not perfect, far from it, but it’s the only big linguistics meeting that has a chance of being a must-attend meeting. And it can achieve that if well-organized sub-communities take a gamble on partnering with the LSA. The meeting is largely as awesome as the people who take part, and it requires good communities to make use of the advantages.

    I like the LSA summer institute, and during my recent spell on the LSA Exec Committee I lobbied repeatedly to have better use made of the fact that so many people are together for a while. But it’s also not something that has any prospect of becoming a must-attend meeting like exists in some other fields.

    As the diversity of linguistics expands, there’s a need to support more and more interesting sub-communities. But there’s a risk that this leads to ever more Balkanization, which is exacerbated when the communities meet separately. And of course that also works against the goal of having people easily affiliate with multiple sub-communities. A solution where strong sub-communities have overlapping meetings would seem to be the most promising solution.

    Reply
    1. Joe Pater Post author

      Thanks for this Colin. I agree completely. I’ve been in touch with Alyson Reed and David Robinson, and they are very supportive of the idea. The only cost to us would be the A/V for the room (about $250 per day), and hotel catering if we want it. We’d get their deeply discounted hotel rates, and all the other advantages you list.

      Perhaps Salt Lake City is too far for people to come for a first meeting (though I’d be interested in combining this trip with some skiing!) but we should definitely keep NYC 2019 on our radar (and it’s New Orleans and San Francisco after that, all places I’d be happy to visit).

      Reply
      1. Colin Phillips

        For what it’s worth, NYC is in 2024, not in 2019. David Robinson jumped on an opportunity to reserve a great rate (though still a little higher than usual) by booking 10 years ahead. The upcoming venues are all fairly attractive, and David has been quite responsive to feedback. SLC is surprisingly reachable (it’s a delta hub), and the city is quite interesting. We chose it based on a student survey that I was involved in, over a couple of other west coast alternatives. Summary of student input: San Francisco is cool, but I’ll have other chances to go there; SLC is intriguing, especially in the winter, I’d be curious to check it out.

        Reply
      2. Joe Pater Post author

        Alyson and David encourage us to speak with members of the sister societies that are already meeting with the LSA: the American Dialect Society, the American Name Society, the Society for the Study of the Indigenous Languages of the Americas, the North American Association for the History of the Language Sciences, the Society for Pidgin and Creole Linguistics, and The Association for Linguistic Evidence.

        I have the MOU that they’ve been using, and that gave me all the info that I needed – e-mail me if you’d like to see it.

        Let me emphasize what I said above – of the American linguistics events we’ve been discussing, the LSA is the only one where co-location wouldn’t mean having to schedule off prime time.

        It still seems to me we may be able to generate more participation amongst computational linguists in our first year by co-locating with the UMass event, but I see going to SLC as equally attractive for different reasons.

        Reply
    2. Thomas Graf

      Colin, do you have information on approximately how many students attend the LSA, in particular undergrads and beginning graduate students? Based on purely anecdotal evidence, my impression is that many students who are still at the beginning of their career submit to and attend NELS, CLS, PLC, and WCCFL, whereas the LSA is something you go to in your 4th and 5th year for networking. And I’d expect that none of these conferences draw as many students as the LSA summer institute or NASSLLI. So if the top priority is to promote computational work to students rather than building bridges to established researchers, that would greatly favor a summer school, no? Of course things would be different if the LSA were to become a must-attend meeting, but I’m wondering about the current situation.

      Reply
      1. Colin Phillips

        @Thomas. I’ve been a regular at the LSA for the past 15 years or so, and my impression is that student attendance is far less focused on dissertators than before. The job market role is largely gone, except for a couple of departments. The number of students is increasing (together with overall attendance), and the student committee has been running a lot of interesting activities (career sessions, networking events, CV & website clinics, publishing workshops, etc.)
        The LSA is still not a must-attend meeting, but it’s the only thing that could hope to achieve that status. The summer institute gets a different sampling of students, generally earlier stage, less likely to have material to present. My sense is that student attendance at institute workshops is patchy, as the meetings have been weakly integrated with the institute. Something that linked a conference/workshop with special training activities would get a larger yield. Traditional department-run conferences are getting smaller and less interesting by the year. People just show up to give their papers, which have little connection to anything else. Meetings like SALT are the exception, in terms of going strong.

        You raise an important issue: if there was a must-attend event in the field, then students would show up regardless of whether they’re presenting, and that’s who you want to reach. But we don’t yet have any conferences like that. This is something that the LSA secretariat would be happy to help with (I wish I could say that the EC is a fount of good ideas, but my impression is that it’s a bit of a conservative body). The addition of mini-courses to the LSA is helping to target early career students.

        Reply
        1. Joe Pater Post author

          @ Thomas – my grad students have started to see the LSA meeting as a venue for their work even from the beginning of their careers. This is different than say 15 years ago. I’d guess that more go there than to the summer institute (in fact, I’m pretty sure).

          @ Colin – should we work at having CUNY, AMP and SALT all co-locate with the LSA? Perhaps CUNY is too interdisciplinary for that. I raised the idea of AMP with Alyson and David, and Alyson pointed out that this might affect phonology submissions to the main session, but didn’t see this necessarily as a deal-breaker.

          Reply
          1. Colin Phillips

            @Joe. Sorry for the very slow reply. I was offline for a while. I think that more co-location of meetings with the LSA meeting would be desirable in general for the field. It avoids Balkanization, it encourages people to interact across areas, and it allows better professional development opportunities for students (and others). And it makes meeting organization less of a burden. The main downside is of course that it lacks the intimacy of small workshops, so it’s not a cure-all. CUNY and SALT probably wouldn’t be ideal targets, for different reasons. We had had some discussions among the linguistic corner of the psycholinguistics world of creating something that might co-locate with LSA.

  15. Tal Linzen

    I’m a bit late to this very interesting debate. The computational *linguistics* conference sounds like a great idea. It would be nice to attract people who don’t strongly identify as “computational people” by co-locating the event with a larger conference like the LSA or perhaps AMP (but making it clear in the call for papers that this isn’t a computational-phonology-only conference – computational *linguistics* is too small as it to be further fragmented into computational *syntax* and *phonology* and *semantics*). I also second the idea of alternating between North America and Europe (maybe GLOW?).

    A few other comments:

    Publication model: I agree with Naomi that it would be unfortunate if the conference ended up luring away the relatively meager amount of (psycho)linguistics work that’s being published in major ACL conferences or ACL workshops like SIGMORPHON, CMCL or CogACLL. And given ACL’s attractiveness for students with a certain profile it is unlikely to do so anyway, at least in the beginning. Some NLP/ML workshops (e.g. the Workshop on Representation Learning for NLP have a cross-submission track where you can present a paper that’s already been published elsewhere (without it counting as a publication at the workshop, of course). We could have a track like that for students who wouldn’t want to “waste” a potential ACL/CogSci publication that might be more useful to their career profile.

    One minor point about co-locating with major ACL conference is that we’d be competing with the literally dozens of proposals ACL/NAACL/EMNLP receive every year, only some of which are accepted (e.g. CMCL was mysteriously rejected this year).

    Online publication: the ACL Anthology already hosts the proceedings of multiple non-ACL conferences, so I don’t see why there should be a problem with getting the conference proceedings hosted there.

    Reply
    1. Joe Pater Post author

      One downside of the summer institute relative to the LSA winter meeting is also that we’d have to compete for a slot. As I understand it, for the winter meeting, we’d just have to form a society and let them know about how many of us to expect. Another downside is that we’d be on our own for accommodation at the summer institute.

      Reply
  16. Gaja Jarosz

    Hi all. I’ve been thinking a lot about all these very useful comments. It seems pretty clear that we have some desiderata that cannot all be simultaneously satisfied. So I’d like to try to clearly lay out what I see as the most central desiderata for this endeavor.

    Highest Priority Goals
    1) we need to attract the core constituents to this conference, especially the first meeting. The core constituents are linguists/cognitive scientists who rely on computational/mathematical approaches and are concerned with questions about the human language faculty.
    2) the conference should be accessible and affordable to linguists, including students. (to repeat from earlier, this rules out co-locating with ACL)
    3) the conference should have quality, peer-reviewed paper submissions. I see this as an important move for the field of linguistics in general, not just this conference. This does not rule out the possibility discussed in the comments above of also having submissions of other kinds, such as presentation-only submissions which have possibly appeared elsewhere.
    4) we want the meeting to be sustainable long-term, with room to become a 2-3 day ‘go-to’ event in linguistics/cognitive science.

    High Priority
    5) Ideally, the conference would alternate US-Europe every other year rather than being solely a US conference to be inclusive of the international community.
    6) Ideally, the conference would be a welcoming/accessible place to linguists who want to learn more about computational/mathematical approaches but don’t (yet) do that sort of work themselves. One way to do this would be to introduce a half-day of workshops or tutorials to initiate the conference. I’m not necessarily proposing this for our first meeting, but something to keep in mind for later.

    Reply
    1. Joe Pater Post author

      Thanks Gaja – this is helpful. I’d like to add another High Priority item that Colin brought up in support of co-locating with the LSA.


      7) Avoiding Balkanization. As we set up specialized conferences, we may contribute to the balkanization of our field (e.g. we may pull computational work out of AMP). To some extent this balkanization is an inevitable consequence of the specialization that is occurring as linguistics grows, but if we can avoid it, so much the better.

      A meeting in early January in a non-tropical location may be far from ideal, but to the extent that we see 7) as important, this may be the only way of addressing this issue.

      Reply
    2. Joe Pater Post author

      Just thought of another High Priority goal:

      8) Increasing the participation of underrepresented groups in computational linguistics

      Given that linguistics as a whole has a relatively high proportion of women compared to computer science, increasing the accessibility of computational linguistics to linguists should increase its accessibility to women. We can more actively pursue this goal by having travel awards for students in these groups – this is something Emma Strubell of UMass CS has suggested for our workshop, and which we could try to do for this conference as well.

      Reply
  17. Gaja Jarosz

    Given the goals above, it seems to me there’s a pretty good consensus on several points:

    – there seems to be pretty uniform support for peer-reviewed paper submissions, possibly published with the ACL anthology. There can additionally be other kinds of ‘presentation-only’, tutorial, and/or poster submissions to maximize the usefulness of the conference for various audiences.
    – there seems to be pretty good support for the idea of alternating US-Europe every year

    The main point of contention seems to be where the first meeting should be held, keeping in mind the long term goals. Although the LSA option looks good ‘on paper’, it really worries me that most of the core contingent responding here in the comments and on the google form say they wouldn’t come if it were held there. While we’d like to attract as much broad interest from the linguistics community overall, I think it’s clear that our primary goal has to be to be attractive to the core contingent or this endeavor won’t be successful. I’m ruling it out yet, but I think I’d need some more evidence that the LSA option is viable.

    I’m also surprised by how many responders seem amenable to coming to AMP 2017 in NYC, given how much general support there seems to be for the conference not being specific to phonology. Is this just because it’s in NYC? And those who are amenable to AMP in NYC, how do you see subsequent meetings playing out (presumably not at subsequent AMPs?).

    The UMass option still sounds quite attractive to me. I understand Giorgio’s point that we don’t want to make the meeting somehow parasitic on a one-time themed event; but I think we can frame the call in such a way as to make it clear that’s not what this is. We don’t have to call it a ‘satellite’ of this meeting; we don’t necessarily even have to officially affiliate with that event. We can announce the first annual meeting of the ‘Society for Computational and Mathematical Modeling in Linguistics’ (or whatever the name will be), and it will happen to be co-located with another event in computational linguistics. The fact that two events in computational linguistics will be happening around the same time in the same place I think will be very attractive. We already have evidence that this would draw in our core contingent: 6/6 people on the form said they would come, and my reading of the comments is that there is quite a bit of interest in this meeting. The timing of the themed workshop is not determined yet, and we would have to coordinate this with all those organizers, but it would probably be possible to get Thursday-Friday for our meeting, if we wanted two days. Friday seems like a pretty ‘prime’ time – not as good as Saturday, but pretty good.

    There does seem to be a good deal of interest here in the comments in the London 2017 CogSci meeting, which is also making me think more about this option. I’d be curious to get more comments indicating whether people would come to that. As Naomi said, it would be nice to have linguistics be a stronger force in Cognitive Science more generally, and it’s nice that there’s a precedent for holding separate but co-located longer conferences. I’m still concerned we wouldn’t get enough linguists to come, but please comment if you disagree! A major drawback of this option is that we would have to apply to be a workshop, and if that fell through, I would certainly not be in a position to organize a conference in London.

    A final option I’d like to address is the option of going totally independent. While I find this option appealing, and it gives us lots of freedom, do you all think there would be enough interest in this to start off with a 2-3 day event for the very first meeting? I don’t have confidence in that but would love to hear your thoughts.

    PS I’m not ruling out the LSA summer institute, but I also don’t see any particularly compelling arguments in favor of it over the other options.

    Reply
    1. Joe Pater Post author

      One advantage of UMass over AMP is that AMP typically runs Friday through Sunday afternoon (e.g. http://blogs.ubc.ca/amp2015/program/), and I expect from conversations with the NYU people that 2017 will be no different. I don’t think we’ll need Friday for our workshop.

      A possible plan is the following. We do a first meeting at UMass in fall 2017, a second one in Europe in 2018 (Giorgio has proposed Paris, which sounds good to me), and then potentially do the third one with the LSA in Jan. 2020 in New Orleans, or at some other location in North America, depending on what people at the UMass meeting prefer.

      Reply
      1. Giorgio

        Thanks Joe. I agree with you that we are indeed getting closer to finalizing a concrete plan. So far, UMass has bid to organize the first meeting of the new computational linguistics conference in 2017, Paris has bid to organize the second meeting in 2018, and then we can take it from there. If UMass organizes the first meeting, obviously UMass gets to decide when exactly to schedule it. I would personally prefer the big new conference to take the entire prime time and the UMass workshop to happen during weekdays, because I hugely care about this new conference and I really would like it to start in the spotlight and with the best possible chances of becoming something big and regular. But well, who am I to tell to the UMass people how they should line up their events!! So, I am personally fully on board with the plan you propose. Let’s go for it! Should we now start a new blog to tackle the delicate issue of the new conference’s name and acronym?

        Reply
        1. Gaja Jarosz

          Giorgio, I’m glad you’re on board with the plan. I understand what you’re saying, but my hope is that the additional appeal of the co-located connectionist workshop will outweigh the advantage of a weekend slot, drawing even more participants than we could otherwise expect. With regarding to scheduling, does this mean that Fall 2018 works for you for organizing in Paris, or are you thinking of somewhat different timing? I think I will start a new post about the name!

          Reply
    2. Naomi Feldman

      Thanks for laying all of this out, Gaja.

      I agree that having presentation-only submission of papers and/or an abstract-based poster session would solve the issue of presenting in multiple venues. This both reduces the urgency of attracting large sections of the ACL/CogSci constituencies, and increases the chances that they’ll attend.

      (CogSci also does give the option of submitting an abstract-based “member only” poster.)

      As far as location, I think my vote would be co-locating with the UMass workshop in the fall. This seems optimal with respect to several issues: (1) funding, (2) motivation for core constituencies attending — the fact that it’s not too distant for our core NECPhon constituencies also helps with that, (3) paper deadline that’s not in the spring, (4) guaranteed acceptance of our co-location request.

      Personally, I prefer weekday conferences (I don’t like to have scheduled work events on weekends), so the timing isn’t an issue for me, though apparently I’m in the minority on that.

      I haven’t seen any compelling arguments to make this phonology-only, so would favor opening it up to all subareas.

      Reply
    3. Tal Linzen

      The main reason I voted for co-locating with AMP was the location; my guess is that international participants will be more likely to attend if the meeting is held in a city that has a major international airport. I might be wrong though – it would be good to hear from more people based outside North America.

      Reply
  18. Jeff Heinz

    Hello, I am joining this conversation late, and I was delighted to see all the comments, and Gaja’s recent synthesis of the goals are ones I agree with along with those in subsequent comments.

    I support the new computational linguistics conference of the sort being promoted here. The venues mentioned earlier by Greg and Thomas (MOL, LACL, FG, etc) are venues along these lines that in my opinion have been under-utilized by linguists working in computational and mathematical approaches to language. In fact, the only difference I can see between the overarching goals of MOL/FG/LACL is that the proposed conference is annual (unlike MOL) and based in North America (unlike FG/LACL). These venues don’t come with high registration fees like the ACL conferences and workshops, though travel to Europe can be expensive. I suppose it’s also the case that phonology has been historically under-represented in these venues because they tend to be dominated by S-type research. While part of me feels that new participation by linguists in the existing venues (both as attendees and in leadership roles) can bring the changes and realize the goals many wish to see, the energy behind the new conference idea is great, so I’ll just emphasize that it’s important that this new conference work in a complementary way with the already existing organizations and communities that have very similar goals.

    Making this conference open to all subfields of linguistics is my preference and it also makes for an interesting counter-force to some types of balkanization (though of course it may lead to others). For this reason, I think co-locating with AMP sends the wrong message. I can see co-location with AMP only if future editions were co-located with conferences focused on other subfields like SALT.

    I am also persuaded by the discussion so far that the LSA is the best place for North American co-location and that has my strongest support. However, I would also support the first event being hosted independently by UMass.

    Regarding tutorials and workshops, these help communities grow. It is my opinion that it is best to make this a whole day affair with each tutorial/workshop lasting at least half a day, if not the whole day. It is not unusual for tutorials in mathematical/computational circles to take this long, and if we want more linguists to become familiar and contribute to these kinds of research efforts, this kind of time is warranted and well worth it.

    Reply
    1. Joe Pater Post author

      Thanks for this Jeff – I’m very glad that you support the general idea of the conference, and am also happy to see that you like the LSA co-location proposal. The main issue with the LSA for the first year at least is that we haven’t got as much evidence that people would attend as for UMass and NYU, and we’ve had some speculation that the existing lineup at the workshop will be a draw that we wouldn’t have at the LSA. For the first meeting especially, it seems important to get as much of the community together as possible. But I am hopeful that we might see our way to meeting in New Orleans for the 3rd year, and continue meeting at the LSA thereafter for North American years if that one goes well. In any case, that’s a discussion for the business meeting of year 1.

      And yes – we should definitely keep complementarity with existing meetings in mind.

      Reply
  19. Matt Goldrick

    Like Jeff, joining this conversation late. Gaja’s summary + Joe’s comment on Jeff seem like the right plan to me: start at UMass, then Europe, then try co-locating with the LSA to see how it goes.

    Reply
  20. Graeme Hirst

    What’s being proposed here seems to be very much like what ACL conferences used to be. It’s certainly true that ACL conferences now have far too much emphasis on NLP, mathematical methods, and feature engineering, and the way to fix this is for all the ideas seen in these comments to be applied to ACL. The NLP crowd needs to learn more about linguistic theory, and the theoreticians need to learn more about NLP. Consider, for example, Joan Bresnan’s LTA talk at last week’s ACL in Berlin, demonstrating how someone who never attends ACL conferences nonetheless has had a profound effect on the field, and vice versa.

    But how to bootstrap this? Co-location may be a good start! Next year’s ACL will be in Vancouver, so those who want an North American venue are in luck. And right now is the time to approach ACL with proposals.

    Reply
    1. Dragomir Radev

      In addition to ACL 2017 in Vancouver, there is also EACL in Valencia, EMNLP IN TBA Europe, and then, in 2018 – LREC.

      Reply
  21. Jason Eisner

    Thanks to Naomi for looping me in. I’m a longtime ACL member.

    I like the idea of running this event on the ACL workshop days. Many ACL folks would be excited to attend. Senior ACL folks remember older traditions in ACL. We like linguistics and often have training in it; we’re sorry we don’t manage to attend linguistics conferences these days; and we think more crosstalk would be useful. Junior ACL folks would attend out of curiosity.

    A new event organized by linguists, with many contributors from outside the usual ACL community, would be perceived as quite different from a standard workshop such as SIGMORPHON. Co-located events do best when they promise something that wouldn’t be found in the main conference.

    So I think it would draw attendance from the ACL crowd if appropriately advertised. We don’t usually get to hear talks by linguists, so that would be fun. I think there could be sharp questions from ACL folks, who will question whether the problems and solutions presented are the appropriate ones.

    I’d be happy to help look for a way to keep costs low.

    Some people were suggesting co-locating with a linguistics conference. I think that would be a pity. It seems that the event is being primarily organized by linguists; so if it is also at a linguistics conference, then it is “by linguists for linguists,” and you might not get any CS folks coming, other than invited speakers. Maybe that would be ok with some of the folks who are leading this effort: the followup post says “The core constituents are linguists/cognitive scientists who …” But I think it would be a shame to miss out on the clash of worldviews. (I’m tempted to offer to write a guest post on the blog about why computational papers by linguists can sometimes be a little frustrating to an ACLer. 🙂

    Reply
    1. Joe Pater Post author

      One possibility that comes to mind is alternating between the ACL and the LSA – right now, I’d say that’d be my preferred option. I’m glad, in any case, that you’ll be at the workshop Jason next fall when we’ll be discussing all of this. It seems like the workshop is still the first choice for our first meeting, given that it looks like it has the best chance of attracting a big crowd for the first iteration, which is very important.

      Reply
    2. Joe Pater Post author

      By the way, the reference to “sharp questions” and an offer to tell us about the weaknesses of our papers brings up another worry about co-locating with ACL: the intimidation factor. My own interactions with you Jason, and with other computer scientists in this area have always been productive and pleasant, but this would be something to worry about in terms of bringing people into the field – I’d really hope we’d make an effort to set the right tone. And in terms of bringing people into the field, we have as one of our goals increasing diversity in CL, and in terms of gender at least, linguistics is way more diverse, so I expect that the LSA participant pool will be more diverse than ACL. Finally, there is a connection between the conduct of discussion periods and diversity, as a very productive discussion on Phonolist has brought out.

      .

      Reply
      1. Jason Eisner

        @Joe – Oh, by “sharp” I meant “incisive” rather than “pointed.” I’m fully on board with respect to creating an inclusive atmosphere, both in person and in written dialogue.

        But I think it’s important for the two communities to respectfully understand each other’s worldviews. Why do papers in the two fields look so different, and what could each learn from the other?

        There was a similar clash of worldviews when machine learning people started talking to statistics people. (It took about a decade of such conversations before CMU created their ML Department in 2006, I’ve heard.) Those two communities were looking at a lot of the same problems, but with different terminology and different values. ML folks were the new kids on the block, and statistics folks thought of them as stealing all the attention by doing engineering work that was practically useful but less scholarly. There is still some of that suspicion among statisticians, I think, but there’s also been a lot of progress in syncing up. At this point, the ML folks generally understand that they are doing computational statistics, and are much more likely to tie into the history and terminology of statistics. And the statistics folks have learned something too, I think.

        Reply
        1. Joe Pater Post author

          Thanks for the reply Jason. I suspected you meant “incisive”, but the ambiguity made me think of the issue, which is good to have out in the open (BTW, Bruce Hayes thanks Alan Prince in his dissertation acknowledgements for “sharp” observations or comments or something, and I’ve always wondered if the ambiguity wasn’t intentional there).

          Reply
  22. Jeff Heinz

    So if the idea is to engage linguists about the importance of computational ideas in linguistic theory then the co-locating with the LSA makes sense. If the idea is to engage mainstream CL about the importance of linguistic theories/questions for CL then co-locating with the ACL makes sense. If the idea is to do a bit of each, then alternating between the ACL and the LSA as Joe suggests makes sense. My understanding is that the idea is to do a bit of both so I would support that.

    Reply
    1. Naomi Feldman

      In principle I support alternating between ACL and LSA. In practice, the LSA is scheduled during a time when I’m often visiting family in South America, and so I wouldn’t be able to attend regularly.

      Another challenge is that LSA is during the winter, whereas ACL is during the summer; so the conference might need to be every 1 1/2 years.

      Reply
  23. Jason Eisner

    As Jeff says, there are multiple good goals, which might not all be on the front burner in the first year.

    Here’s a possible plan for an event that aims to mix folks from both communities, even if it’s not the first event:

    Full-length submitted papers, with explicitly mixed reviewing (e.g., 2 linguists + 2 computer scientists per paper).

    A longer-than-usual period to revise the paper in response to reviewer comments. And/or a post-proceedings (this could probably be arranged with the ACL Anthology even though it’s not standard).

    In addition, because this would be a kind of summit between two communities, there should be explicit bridging attempts. Getting the right topics and people for invited talks and panel discussions would be crucial.

    Since the two communities haven’t always followed each other closely, I think there should be some way to bring each other up to date. For example, there could be a poster session for PREVIOUSLY published work. Such a poster session could be on the large side (in fact, you want enough posters to prevent too many people crowding around each one). But it should probably be refereed — again with mixed reviewing — to ensure that the most interesting and topical work was being presented and that there was a balance of topics and perspectives. So if my group wanted to present a poster about our work on recovering underlying forms, we’d have to send in a copy of our TACL paper, or perhaps a 4-page synopsis of that paper and its followup papers. Such a synopsis would appear in the proceedings, similar to the AAAI Nectar Track (http://www.aaai.org/Conferences/AAAI/2011/aaai11nectarcall.php).

    Conceivably there could be a forum for comments on accepted papers, with the anonymous reviews being the initial comments (http://openreview.net/, http://deeplearning.net/2015/11/26/open-discussions-for-iclr-2016-is-now-open/). But that might be too much novelty at once. 🙂

    As I said above, I think co-locating with ACL, NAACL, or EMNLP would be the best bet for getting the NLP folks to attend. The NLP folks are generally open to bringing their toolbox to bear on new problems. They have a good toolbox for modeling, algorithms, and experimental methodology. So getting them excited about linguistics questions again seems like one worthy goal.

    Reply
  24. Brendan O'Connor

    Hi, on the submission issue — for the NLP+CSS workshop this year (EMNLP 2016) we have an explicit non-archival option designed to be more compatible with publishing practices in social science and linguistics. https://sites.google.com/site/nlpandcss/nlpcss-at-emnlp-2016

    This dual-track submission style may be (or ought to be?) cropping up more in CS + interdisciplinary meetings. ICWSM added it a while back (it’s technically under the AAAI umbrella but has considerable psychology, comm, sociology participation).

    Reply

Leave a Reply to Gaja Jarosz Cancel reply

Your email address will not be published. Required fields are marked *