In our previous post, we hosted a discussion about a new conference for linguists and cognitive scientists on computational and mathematical modeling. In this post I’d like to solicit comments and suggestions about possible names for the conference. Before I lay out some existing suggestions for commentary, I’d like to summarize the overall plan and goals for the conference that emerged from that discussion:
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 Goals
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.
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.
8) Increasing the participation of underrepresented groups in computational linguistics.
Overall Tentative/Consensus Plan
1) The first meeting is to be tentatively held at UMass in Fall 2017 in conjunction with a one-time workshop on computational modeling of language (invited speakers, pending funding, include Jacob Andreas, Emily Bender, Sam Bowman, Chris Dyer, Jason Eisner, Bob Frank, Matt Goldrick, Sharon Goldwater, and Paul Smolensky). The exact schedule is unknown at this point, but tentatively the new conference may be on a Friday or a Thursday-Friday, with the workshop probably Saturday-Sunday.
2) The second meeting is scheduled to be in Paris in Fall/Winter 2018, organized by Giorgio Magri.
3) We will have a general discussion of hosting options for subsequent meetings at the first meeting at UMass. One prominent possibility is holding the third meeting in conjunction with the LSA annual meeting in New Orleans in Jan. 2020.
4) The current plan is still to have paper submissions, possibly published with the ACL anthology (though stay tuned for another post to discuss this further).
Ok, so on to the candidate names! I think the current favorite in offline discussions among us is “Computational and Mathematical Modeling in Linguistics” with the acronym CAMML or CAMMIL or maybe even CAMMLing or CAMMILing. What do you think? I like that it is clearly about linguistics and that it is inclusive of both computational and mathematical approaches, and that it has a cute and pronounceable acronym. Earlier variants had “Linguistic Theory (LT)” or “Theoretical Linguistics (TL)” in them (like CLINT, CALT, CAMLT, or CATL, etc), there is also the option to add “meeting” (M) or “annual meeting” (AM) or “Society” (S) or “conference” (C) somewhere (yielding things like CALM, AMCTL, SCATL, etc). I’m sure there are many other possibilities, but I will leave off here with my favorite: CAMML (or is it CAMMLing).
Thanks Gaja for summarizing the preceding discussion and for starting this new blog! I like very much the proposed name “Computational and Mathematical Modeling in Linguisticsâ€. Perhaps, replacing ”Linguistics” with ”Linguistic theory” (or ”Theoretical Linguistics”) might make it even better, because it might provide a better characterization of what the conference is about. The resulting acronym CaMMiLT (CaMMiTL) is a bit long but not too bad after all.
I disagree — I like “Linguistics” — as it more clearly includes interdisciplinary areas like computational psycholinguistics.
I have the impression that, if a conference name isn’t one (obviously pronounceable) syllable, three-letter acronyms are much catchier (LSA, ACL, SfN). Could we not just keep it minimal? Say, MCL (Mathematical and Computational Linguistics )? Or a permutation. CML, LMC.
The conference will get its niche from the writing of the first few calls, from the attendance of the first few conferences, and from the vibe – not from its name. We risk turning this into a game of Twitter (fit everything you want to say in X characters).
I wanted to expand a bit on the rationale for my comment above.
I’ve shared the original blog post with a bunch of people here at CogSci, and there’s been some discussion about it. The comment above stems from a conversation I had with Roger Levy. He raised the concern that this conference could provide a new way for linguists to wall themselves off from other computational researchers who study language, defining narrowly what counts as linguistics. I can identify with this concern, as the biggest hurdle I had to cross when getting a linguistics job was convincing people that I was a “real” linguist. I’ve always found NECPhon to be a welcoming environment, and would like to make sure that reviewers and audiences also don’t have an attitude of trying to weed out anyone who isn’t a “real” linguist.
“Linguistic theory” and “theoretical linguistics” are phrases that are often used to create such boundaries. So while I don’t care that much about the conference name, I think including those phrases has the potential to do real damage.
On another note: Lila Gleitman just won the Rumelhart prize (long overdue!), so there will likely be more language researchers at next year’s CogSci than usual.
I agree with Naomi on this. I was already worried that “linguistics” rather than “language” might even set the conference up as more exclusive than it needed to be. For me, “linguistics” is a good compromise between the perhaps excessive breadth of “language”, and the perhaps excessive narrowness of “linguistic theory”.
Right, I agree too. I think “linguistic theory” gives an impression of being overly narrow, and we want to be inclusive to computational psycholinguists and other groups that work on human language who might feel excluded by that phrase. Is “linguistics” neutral enough? I would prefer that over “language” if it is, but I could go either way. Of course, the call and so forth will be more significant in defining the relevant community, but I think these discussions are useful to have for both.
I also would prefer to set up a big tent, but I think it also means making it extremely clear that linguists are invited, and that there will be linguists there, that theoretical work can be accepted, that the audience will understand the issues, and that they won’t be hostile. There really isn’t a big tent conference like this, and so there isn’t any reference point. If the question “do other linguists go to this” isn’t properly addressed, this won’t work.
I’d think “linguistics” should be in the title. After all, that never stopped the ACL from becoming what it is (not linguistics). I think “theoretical,” given the constraints of titles to be basically conjunctive, won’t work (i.e., it doesn’t *have* to be theoretical).
The concern raised by Roger via Naomi is a real one with which I sympathize too. Computational researchers interested in language come from many traditions including psychology, cogsci, different branches of theoretical computer science (logic, automata, information), NLP and so on. So I am also interested in how the organizers’ will (in Naomi’s words) “make sure that reviewers and audiences also don’t have an attitude of trying to weed out anyone who isn’t a ‘real’ linguist.”
As for the name, “language science” is a term people have used to move beyond “linguistics”. What about “Computational Language Science”? While it is unfortunate this acronym overlaps completely with “Chicago Linguistics Society” I still like it because is is succinct and to the point. A longer alternative is Computational Language Science Research (CSLR).
What about Computational Science of Language (CSL)?
Oh, and not “walling themselves off” isn’t just about letting others in (and encouraging them in), it’s also encouraging linguists to bring their perspective to these other communities, which means participating in them. So part of the success of the conference series and its mission will in part be measured by an increased participation in *other* related venues by students and faculty from linguistics departments.
I agree with the concerns raised here; I think “Language Science” is a useful neutral term. I guess a question is whether those working in more generative frameworks also think it is neutral. Hopefully people read it as Jeff intends (that’s what I mean when I use it!): “encouraging linguists to bring their perspective to these other communities.”
I’m without a real keyboard, so I can only make a quick response, but in response to Matt’s question I do not think explicitly trying to avoid the word “linguistics” is a neutral action at all (there probably isn’t a choice that is both neutral and meaningful), and I would strongly advocate using the word “linguistics” in the conference title or description.
I suspect reactions to this this may differ depending on ways in which particular subfields are differently polarized w.r.t. computational work.
Another way of signaling our shared interest would be to use the term “Linguistic Cognition”, which seems more informative than “Language Science”, and less discipline-specific than “Linguistics”. I don’t have a suggestion for a cute acronym using it though…
Thanks everyone! I also just want to add that if you have any additional comments about other aspects of the tentative plan (especially if we haven’t heard from you), please feel free to continue to comment on them too. I’d like to gather as much information as we can to make sure this is successful!
Quick comment on name and inclusiveness. In my experience, the scope and the vibe is defined by the people who collectively lead the effort. If the right range of people is there, and if they set the right tenor, then the name won’t matter. If people perceive that their interests are a complete outlier, or that their concerns or approach are sneered at, then you’ll have lost them. One of the things that has worked reasonably well about the CUNY conference over the years is that it has worked to keep a biggish tent. This is helped by an informal but active steering group that consists of everybody who has ever organized the meeting.
Thanks, Colin. I totally agree – that will be a really important decision.
I’m not necessarily opposed to them, but I’m worried that Computational Language Science or Computational Science of Language sounds rather vague and amorphous and might alienate the theoretically oriented linguists. I was also just thinking how it’s frustrating we can’t just use computational linguistics because ACL is already using it, but is that really true? Can’t we also use it? How about Computational and Mathematical Linguistics (CML)? or Computational Linguistics Meeting (CLM)? Or Meeting on Computational Linguistics (MCL)? or Computational Linguistics Conference (CLC), or Society of Computational Linguistics (SCL)? I like CML the most of these.
I like CML.
The triliteral root C-M-L had some support already. It includes “computational”, “mathematical”, and “linguistics” (there seems to be some doubt over using this as opposed to “language science” – I can’t make my mind up either – but people seem to prefer either to “language” – maybe implicitly, the desire is to narrow the scope to exclude or disprefer strictly applied work; so, simmering in the background is another iteration at some point over what exactly the scope is – but it seems like “linguistics” is fine anyway).
I also like short, as noted, and CML is also my favourite among the three-letter acronyms. It’s nicely pronounceable as (1) see-emm-ell, (2) camel, and (3) Camille, and lends itself in the future to such clever collocated workshop names as OCML, CML-OT, and CML-ATATURK.
+1.
According to ACLweb.org: “What is computational linguistics? Computational linguistics is the scientific study of language from a computational perspective. Computational linguists are interested in providing computational models of various kinds of linguistic phenomena.” So in that sense, we should call ourselves simply CL 🙂 Unfortunately, it won’t work. But I like anything that is similar to “computational and mathematical linguistics” (using any metrics).
Like most of the other commenters I agree that it would be best to avoid terms that are associated with a particular subset of the computational linguistics community, such as “theoretical linguistics”, “the language faculty” or even “cognitive” – there’s interesting historical linguistics work in the ACL community that doesn’t necessary take a cognitive perspective.
A couple pronounceable acronyms: CALS (Computational Approaches to the Language Sciences); CASL (Computational Approaches to the Study of Language).
I like Gaja’s suggestion of “Computational and Mathematical Linguistics”. I agree that CSL and CLS are too amorphous – they don’t signal what we are trying to do with this conference. In particular, it’s not clear from those names that this is not an AI-oriented endeavor. I’d like to take back my “Linguistic Cognition” suggestion, not only because of Tal’s point, but also because I worry that it might be taken as indicating that it is more exclusively focused on computational psycholinguistics than it is.
It’s kind of hard for me to imagine that there is a group of people who would want to participate in this conference who would be put off by the use of “linguistics” in the title (note that Naomi, who raised the issue, was OK with it). If they are put off by that name, then they are also going to be put off by the content of a lot of the talks, if the people participating in this discussion present their work. It’s true, as Ewan says, that the ACL has become AI-oriented even with their name, but I think that’s a case of the name no longer suiting the conference, insofar as a most of the work presented at the ACL is more engineering than linguistics. Finally, I think it’s worth bearing in mind the goal of more firmly establishing CL as a sub-discipline within linguistics – this is a big reason why I’m interested in this. I think having linguistics in the title may help a little in this regard.
CML and CAML are unfortunately not good for googling (and there is the language CAML). As Michael Becker pointed out on Facebook, CAMML is google-able, and while this may seem a trivial point, I have found AMP do be a bit of a pain in this regard, since there are too many competitors. So I suppose I’m back to supporting Gaja’s original suggestion.
And yes, everyone who says that the name is not the prime determinant of what the conference ends up being is right. But I think this discussion has been independently useful for bringing up people’s ideas of what this conference should be.
I also like Computational and Mathematical Linguistics, and wouldn’t worry too much about googleability. “ACL” is not googleable, but “ACL linguistics” is, and people generally don’t have a problem finding the ACL’s conference website.
You’re right, and CML is more elegant.
I’m convinced, +1 for CML.
I think that ”Computational and Mathematical Linguistics” does not properly characterize the envisaged conference. It sounds to me like a subsection of the ACL conference with a stronger focus on mathematics!! I prefer Gaja’s initial proposal way better, because I think it better describes the spirit of the conference. If we want to go for something shorter, what about ”Computation and Linguistic Theory” (CALT)?
Giorgio, I don’t understand your objection to “Computational and Mathematical Linguistics” as opposed to my original suggestion of “Computational and Mathematical Modeling in Linguistics”. All the same key phrases are there, all that’s missing is “modeling in” – I don’t see how that changes the implied spirit of the conference.
My intuition is that “Computational and Mathematical Linguistics†sounds like the union of ”Computational Linguistics” and ”Mathematical Linguistics”. Unfortunately, those two expressions are already taken: the first has come to denote the research represented by the ACL and the journal ”Computational Linguistics”; the second sounds to me like a a synonymous of ”Formal Language Theory”, a very formal subfield of Computer Science. Of course, we want to interact with and borrow tools from both communities, but I do not see the new envisioned conference as *representative* of neither of those two communities. That proposed name thus seems to me too broad.
Indeed, my intuition is that there is a huge difference between ”Computational Linguistics” (which, to repeat, has been taken by the ACL) and something like ”Computation and/in Linguistics” (which comes closer I think to what we all have in mind).
And — just to reiterate my earlier point — ”Computation and Linguistic Theory” sounds to me even better. That title seems to me a good shorthand for the title ”Computation and phonology, morphology, syntax, semantics, language processing, historical linguistics, etcetera”, as the latter subfields are aptly summarized by the expression ”Linguistic Theory”, as opposed to things like ”text summarization, email spam classification, automatic translation, etcetera”, which I feel as having less to do with linguistic theory.
Especially if the new conference turned out in the end to be co-located with the ACL (although I fully share Tal’s and Naomi’s concerns below about costs), I think that a title which contains ”Linguistic Theory” would be particularly apt to promote the effort of a bunch of linguists to reach out to the ACL community.
I don’t mind “Modeling in”, but do have an aversion to “Linguistic theory”, because I do not think those outside of what’s often referred to as “Theoretical linguistics” would see this term as being open or inviting.
If we co-locate with ACL, however, I have no problem with “theoretical” or “theory” being in the title — and then I would just submit to the psycholinguistics track at ACL instead of to this workshop.
I think there’s been a lot of good arguments against using ‘linguistic theory’ or ‘theoretical linguistics’ already. Another one I haven’t heard mentioned, is that it can give the impression that this conference is only interested in submissions that directly connect to or work on some particular linguistic theory. I think there are plenty of computational questions that would be of interest to this conference that are positioned more broadly than that, and I wouldn’t want to discourage these submissions.
As a compromise, how about we use the name “Computational and Mathematical Modeling in Linguistics” and the acronym CML, which fits both?
I am actually fully on board with “Computational and Mathematical Modeling in Linguisticsâ€. It does seem to me a very good option and it has been met with approval on the blog. And I see no reasons not to go with the full acronym (CaMMiL or variants thereof). I am afraid the shorter acronym CML will leave people wondering whether it is ”Mathematical” or ”Modeling” which was left out from the acronym!
I posted my comment below before reading Joe’s post here. I completely agree with Joe’s point. If not ”linguistic theory”, at least ”linguistics” should appear in the name of the conference
I agree with Ewan that, in the end, the phonotactics of the acronym might be just as important as the semantics of the name. Yet, let me elaborate on why I think that something like ”linguistic theory” (or at the very least ”linguistics”) should appear in the name of the new conference. It used to be the case that computational work had essentially no place within linguistics: you would see none at linguistics conferences, linguistics department would not teach any nor try to make any connection with computational people. Things have changed tremendously within linguistics in recent years–and in particular within phonology (this might indeed turn out to be one of the most important legacies of constraint-based phonology). Things have changed so much that we (when we started to chat about this new conference over lunch at Amherst) thought that there might really be the beginning of a new subfield of linguistics, next to other traditional subfields like syntax, semantics, and phonology. And we thought that, in order to highlight this fact and contribute to this trend, it was a good idea to put together a new conference dedicated to this subfield, on a par with other linguistics conferences such as SALT and AMP.
I think that using ”linguistic theory” in the name of the conference is very important to send out the crucial message that we want to send out: that linguistics has evolved substantially and has developed a new, strong, deep interest for computational and mathematical modeling. I don’t see why this message should keep people away. All to the contrary. This key message should crucially help to attract people by signaling the shift in attitude within linguistics.
To me, it is pretty clear that this conference is a conference organized by linguists for linguists. Just like SALT and AMP. When AMP was started, no one worried that the word ”phonology” implied we were walling ourselves off. And no one suggested that we should have replaced AMP with something like ”annual meeting in the science of linguistic sounds”. As another example, SALT (which stands for ”semantics and linguistic theory”) is a great conference of the highest level. For the last five years or so, one of the four invited speakers at SALT has always been a psycholinguist. Indeed, formal semantics has lead to an incredible amount of experimental psycholinguistic work of the highest level in recent years. I really do not think that the semanticists who go to SALT feel in any way walled off from psycholinguistics. And none of them feels that the name of their conference is too restrictive.
We are linguists. We are very interested in what is happening outside of linguistics. We obviously want no walls around our discipline (irrespectively of who is being walled in and who is free outside). But we do linguistics, and we organize linguistic conferences. So what is wrong with making that explicit?
Hi all,
I am a late-comer to this discussion, but I am keenly interested in the ideas being discussed (thank you to Naomi for conveying the initial concerns I raised in conversation). I would like to make the case for taking advantage of mechanisms that are available through ACL, and in particular considering (1) collocating a conference of this nature with a *ACL (i.e. ACL/NAACL/EACL/etc.) meeting, and (2) mutual encouragement among the members of this community to also submit papers to the *ACL meeting with which the conference is collocated. It seems to me that this approach would have many advantages:
* The logistics would surely be easier, from paper submission & reviewing to venue arrangements to publication of proceedings — ACL has this all worked out pretty well.
* It would bring more linguists into participation in/contact with *ACL events, which, as Jeff points out, would promote bi-directional intellectual exchange between this community and other *ACL attendees.
* Conversely, it would minimize balkanization and fragmentation into different sub-communities of computational research on language. Working with the ACL instead of separately from it would be, in Colin’s words, an unmistakably “big-tent” approach.
* It would be likely to increase participation by researchers who already regularly attend *ACL meetings, who from a travel-commitment perspective would be able to get this conference as a “freebie” (surely most of us would like to attend more conferences than we have time for!).
My impression from the numerous comments on this and the previous post is that the main objections to *ACL collocation are that (1) ACL is not interested in non-engineering-focused work; and (2) the registration fees for ACL workshops and collocated conferences are prohibitive, especially for students in linguistics. I actually think there are good arguments against these objections:
* Although engineering-focused work clearly dominates *ACL, in my own estimation there is no bias against what I’ll call without intended prejudice more “scientifically focused” work in the review and paper selection processes at these conferences. In fact, my impression aligns with Robert’s that the leading figures in ACL generally want to see more scientifically focused work at the conferences, and so if anything review & paper selection may be more favorable to this type of work than to engineering work! And I think this trend is strengthening in recent years. I remember when in 2010 ACL had a Psycholinguistics submission track for the first time (at least since I entered the field as a graduate student), and how striking that was. These days, it seems that there is a Psycholinguistics or Cognitive Modeling or Linguistic Aspects of CL track at at least one *ACL conference every year, and sometimes at several. I think that a concerted effort by the members of this community to submit their best work to *ACL could easily make this type of work an even more substantial and regular *ACL fixture.
* Students authors and co-authors of *ACL conference and workshop papers are eligible for student travel grants, which waive all registration costs and also provide travel funds.
* It might be possible to defray costs (e.g., for invited speakers — and perhaps reducing registration costs?) through NSF conference/workshop grants. An analogue: there is a strong history of the NSF funding *ACL student sessions.
More generally, I would echo Robert’s suggestion that leading ACL figures be engaged more extensively in this conversation. I think that many of them would rightly see considerable potential value in closer engagement with a wide variety of linguists, and everyone would stand to gain from open discussion of the possibilities.
Hal just forwarded links to these blog posts to the people in charge of NAACL and ACL.
Essentially, the idea would be that instead of starting a brand new independent conference, we get together and try to revamp SIGMorPhon. We would replace it with something broader in scope, namely not restricted to phonology and morphology and thus make it the new conference that we actually wanted to organize.
I am personally very sympathetic to this idea, as I definitely see the advantages that you listed and I agree on their importance. Also, it seems that the issue of registration cost has gotten better over the years. The last time I attended SIGMorPhon was 2014 and I was charged a staggering 355$ in registration fee. But I see from the registration page of the ACL 2016 (http://acl2016.org/index.php?article_id=52) that the cost of registration for a 1-day workshop without main conference (I think this option was not available in 2014) is now 230$ for regular early registration and $160 for student early registration. That is still a little bit more than a plain linguistics conference, but it looks doable.
I just got back from ACL 2016. I was going to write a longer post on my experience but for the moment I’ll summarize a couple of points here.
There were a few linguistic-y papers, and people quite liked them (one of them, which Roger was incidentally involved in, got the Best Paper Award). Most of the action in the conference was still on the engineering side, of course, and (psycho)linguists were a relatively small fraction of the 1600 attendees of the conference.
As for the registration fees, the cost that Giorgio is citing is only for workshop registration. As a post-doc I paid a total of $875 for obligatory ACL membership, main conference registration and workshop registration. In the past my single attempt to obtain a student travel grant (for EMNLP 2015, when I was paying my own way as a linguistics PhD student) was unsuccessful.
I am also in favor of trying to find a way to co-locate with ACL. In particular, the goal of decreasing balkanization within computational approaches to language seems to be the flip side of the argument against balkanization in linguistics, and very worthwhile.
However, for co-locating with ACL to be a real possibility given the goals of this conference, we would need to find a way to lower the registration fees to somewhere on the order of $100 (total) for students, $200 (total) for non-students, for what would ideally be a two-day workshop eventually. And I’d like to see attendees at this conference also be able to attend the main ACL conference, so ideally those registration costs would be lowered to more manageable levels as well.
I’m less optimistic than others who have posted here that this is a real possibility, as I’ve already raised this issue several times with the ACL leadership (i.e., the issue that high conference fees are the main barrier to participation by linguists) and have not yet seen efforts to address it. Even simple one-off requests — e.g., asking for volunteer slots for local linguistics students when ACL was in Baltimore, to enable members underrepresented groups whose research was in computational linguistics to participate in the conference — were declined.
I’m hoping that this discussion can motivate the ACL/NAACL organizers to address the issue of financial barriers to participation in a serious way.
I see: I forgot to add the 100$ for mandatory ACL membership fee. So we are up to $260 for student registration and 330$ for regular registration, and that would be only for the one-day event, not for the ACL main conference. That is a lot of money indeed for a linguist’s budget.
It sounds like a great goal. I’d vote to work tentatively towards doing it, but not right away.
The cost must be substantially less than the cost of the LSA, I would think, to make it not a problem. It would be worth taking some time to sort out if there’s a way to subsidize it, or put it in an adjacent venue to defray the cost. If that doesn’t happen in the first couple of years I think it would be better not to attach it for the time being.
It’s also worth thinking about: how exactly are we going to target “normal” (borderline/curious) linguists to get them to attend? If the first/second iteration of the conference is at the ACL it might be (wrongly) perceived to have some prohibitive applied tinge, in spite of the best urging of the organizers. Walls take longer to take down than to put up. Once the conference is established, that wouldn’t be an issue.
There are many worthwhile goals to have for conferences, and I see the appeal of reducing balkanization within computational linguistics/NLP and getting more linguists to attend ACL. However, those were not the motivations for this conference, and I’m afraid I see these desiderata as contradictory to the main goals for this conference, at least for the time being given the current situation. The main motivations for this conference are to make computational linguistics more accessible and more central within Linguistics. It was indeed the failure of other ACL-associated workshops, particularly sigmorphon (which I am most familiar with) to attract linguists that motivated this effort in the first place. That, together with the observation that there are growing numbers of linguists (perhaps esp in phonology) who rely on computational approaches, attend NECPhon (for example), but have never been to or would consider going to ACL.
Yes, there is a small community of computational *linguists* who are already deeply involved in computation and occasionally or regularly attend ACL. There is a much, much larger and growing community of linguists who use, want to use (or just want to hear the results of others who use) computational approaches applied to linguistic questions. I see the motivation behind this conference as making computational linguistics into a more coherent and centralized movement within linguistics, and also to make a ‘bridge’ to the broader computational linguistics community (including ACL). ACL is simply inaccessible to most linguists, and I just don’t see how we can meet the primary goals of this conference by associating with it. We can debate the reasons why linguists don’t go to ACL (cost, applied/engineering focus, technical background assumed), but the fact is they don’t. And we want this conference to be for linguists. Maybe once this movement is strong enough, this ‘bridge’ could be moved closer to ACL, that would be great. But I think the first step in getting more linguists involved is building this bridge from the linguistics side.
It also occurs to me that part of what’s contributing to the linguistics-ACL divide is that the area within linguistics that’s seeing the largest growth in application of computational methods, phonology, has no representation whatsoever at the ACL. For that to change, I think someone will need to show and convince the ACL community that phonological theory can actually contribute something to mainstream NLP problems like ASR or MT. I read a paper last year that did this in a specialized domain, which I got very excited about, but as far as I understand, we are very far away from establishing this in general.
It seems like almost everyone thinks this is impractical, and I agree that for now it is and, based on what Naomi says, maybe forever more, but to articulate a bit better the motivation for this, in defence of the general idea:
The ACL has *become* an applied conference and has *become* swamped with applied people with only small bastions of, as Roger put it, “scientific” work (SIGMORPHON, CMCL). But this isn’t the way it has to be. That’s not what the ACL was founded for, that’s not the way it used to be, and, at least several years ago, there was something of a trope at ACL keynotes to bemoan this rift (at least, I saw one at NAACL, it referenced one the year before, and long before that I’d heard tell of another).
Similarly I know Jeff has for a long time tried to insist that we have to work to “make ACL great again” by putting the emphasis on the L, something like this. So there’s a widespread desire for rapprochement, but there are barriers, it doesn’t happen. So here’s an opportunity to do that, which is good for all the reasons everybody says above, and maybe has the chance to finally make a real difference.
I think it’s premature and I would prefer your strategy of, let *this* conference grow strong in its own right, give ACL a *reason* to want to co-locate with *it*.
But I would defend this idea as being a solid one.
Just to correct the record, it is not the case that phonology has no representation whatsoever at ACL. Just to point to a few examples, Jeff has a bunch of papers published in *ACL conferences. My former students Gabriel Doyle, Klinton Bicknell, and I published a paper on non-parametric OT learning in ACL 2014. There is an interesting paper by Johnson, Pater, Staubs, and Dupoux in NAACL 2015 on a theoretically crucial issue in MaxEnt OT. In fact, both NAACL and ACL had a phonology(+morphology) session in each of the last two years:
NAACL 2015: Phonology, Morphology and Word Segmentation
ACL 2015: Morphology, Phonology
NAACL 2016: Morphology & Phonology
ACL 2016: Phonology and Morphology
(Though it should be noted that neither EACL nor ACL had a phonology session in 2014.)
Thanks, Roger. You’re right, ‘none whatsoever’ is too strong, my bad. Despite these promising examples, my sense is that there is not enough there to get on the radar of most phonologists, even when SIGMORPHON is included. I’m not sure how or if this can be changed. If SIGMORPHON couldn’t do it, it’s hard to see how another ACL workshop would be different.
At the risk of basically repeating what Gaja and others have already said, here’s my two cents.
I think that Roger’s idea of co-locating a conference of the general nature that we’ve been talking about with the ACL is great one. Maybe SIGMORPHON and related ventures could join forces, and create an umbrella suborganization within ACL?
But I’d like to come back to something Jeff said a while back – that this conference should *complement* the existing endeavors. That’s the way I see it, as something different from what has been, and could be, done within the ACL. The idea is to make something maximally accessible to linguists, and in the long run, I still think that co-location with the LSA will be the best way to accomplish that.
This is not “walling off”, this is shoring up one side of the bridge. Linguists need to better understand the importance of computational and mathematical modeling to answering the questions that they are interested in. That understanding is key to increasing the number of computational posts in linguistics departments, and to establishing CL as an intellectual subdiscipline within linguistics (it’s too often seen as intellectually outside, or just methods). In addition, recruiting young linguists to this subdiscipline requires an accessible forum. Even if we somehow manage to get the fees to the ACL down for this sub-conference, it’s still a distinct conference from the LSA, and there will be less interaction with the general linguistics community.
Part of the tension here is that those of us participating in this discussion are a diverse bunch, with different pre-existing commitments in terms of the other conferences we need to attend, and different ideas about what we’d like this conference to be. It’s possible that we are in fact talking about two separate initiatives: one within the ACL community, and one within the linguistics community.
I’m very much in favor of whatever can be done to increase communication across these communities (which are of course not completely distinct, as this conversation shows). The workshop that we’re discussing co-locating with in the first year has that as its basic goal (the specific theme is neural nets and linguistic structure, aiming at productive discussion, rather than paradigm battling). This goes back to something Ewan raised a while back – would we want “applied” work at this conference? I’d vote yes, so long as it would be of interest to linguists (whatever we’d take that to mean).
Let me say that this discussion is very different from anything we talked about in setting up AMP (Annual Meeting on Phonology). With AMP, we had a discipline we were targeting. Here, it feels like we’re trying to figure out what the field is that the conference is targeting, or maybe even creating a field!
Name that conference:
Let me suggest Computational Answers to Linguistic Questions (CALQ), which clearly delineates the scope.
Computational methods in linguistics are often CALQues of their original versions in machine learning or NLP, or vice-versa. And of course they’re theories whose predictions must be CALQulated.
I’m also ok with Computational and Mathematical Modeling in Linguistics (CMML) — but that title wouldn’t jump out in a list of ACL workshops. Too similar to “computational linguistics” and “mathematics of language.” I might not realize “ooh! new! with actual linguists!”
p.s. I just posted a couple of comments about co-locating with ACL, at the previous post http://websites.umass.edu/comphon/2016/07/28/conference-on-computational-approaches-to-linguistics/ .
I agree that making the scope of the conference/workshop as transparent as possible through the name is a worthy goal, though this discussion has made it clear that this endeavor is somewhat complicated by the fact that we’re still trying to work out the scope! CALQ seems pretty good, though I wonder whether we might also sometimes get Linguistic Answers to Computational Questions.
I was reflecting on the fact that as I said in an earlier post we are essentially carving out a sub-discipline, and wondering if we might be able to come up with a conference name that also serves as a sub-discipline name (for e.g. job ads, a journal name…), like Laboratory Phonology managed to pull off. My best candidates so far are:
Linguistic Computation
Computation in Linguistics
Computational Approaches to Linguistics
It seems to me the sub-discipline you are carving out has a long history which has overlapped with linguistic theory to greater and lesser extents over time. “Computational Linguistics” seems the best name for this sub-discipline to me.
The LSA is looking to develop Special Interest Groups (SIGs).
http://www.linguisticsociety.org/content/special-interest-groups
If that is still an option, why not make this a LSA SIG in CL?
Then the name matches the sub-discipline, it is clearly in the purview of linguistics because it is inside the LSA, and thus it is also clearly distinct from the ACL.
Whatever the conference is called, I do agree that CL is the appropriate name for the sub-discipline:
https://www.quora.com/How-is-Computational-Linguistics-different-from-Natural-Language-Processing/answer/Jason-Eisner?share=1
CL is already taken as a journal name, of course.
Thanks Jeff. The definition that Jason gives in the Quora reply does indeed sound like what we are talking about. I was imagining we were talking about a subpart of CL, but I guess that’s because I’ve been too influenced by the contents of the ACL.
I hadn’t heard about SIGs before. I don’t think that’s quite what people have in mind with this endeavor, but I could be wrong.
Sorry all, I missed this discussion until yesterday.
Sounds like this ship has sailed but here are a few notes about ACL colocation.
– I think a nontrivial amount, though yes, perhaps a minority, of the ACL community would appreciate such an event being colocated with ACL. On the other hand, if the central purpose is to target psychologists and linguists, presumably colocation with ACL has little benefit since there are so few would be planning to go already.
– Mechanically, one of the co-location options is to be classified as an ACL workshop, starting the day after the ACL main conference, with separate registration. Registration fees (without ACL) are in the $100-$200 range. (Another option for a more well-established meeting might be to be officially colocated conference like CoNLL (or *SEM?); I don’t quite understand the details of this. CoNLL is a two-day conference that runs at the same time as ACL workshops.)
I remember the ACL visualization workshop in 2014 had many attendees (was it a majority?? based on a quick show of hands) who were there only for the workshop, not ACL itself.
Of course, investigating the low burdens of being colocated to ACL is silly at some point since it’s admitting the low benefits of colocation in the first place.
Oops to correct the registration numbers: maybe more like $250-$400
http://acl2016.org/index.php?article_id=52
Now I see Jason Eisner posted a bunch in the other thread on this http://websites.umass.edu/comphon/2016/07/28/conference-on-computational-approaches-to-linguistics/#comment-1003
ok never mind i’m way behind, looking at the other thread now
Holding such a conference right before or right after ACL (or NAACL), in the same or nearby location, without being part of the ACL conference, sounds like a great idea to me.
Can I just say how happy I am to see this development. It is wonderful to see linguists wanting to make greater use of computational tools to better understand their very real problems.
Could I note that the LSA Summer Institutes have, for many years, attempted to provide linguists with skills in computational tools and the encouragement to use them (although maybe the latter gets diluted when students return to home). In the upcoming (2017) LSA Institute, besides Giorgio Magri’s course in Computational Phonology, Mark Steedman will be giving an introductory course for linguists in CCG (Computational Categorial Grammar, and Julia Hirschberg, a course in Intonation and Computation. (This is in addition to Sandra Kuebler’s course on Computational Linguistics for linguists, and Nathan Schneider’s course in Corpus Linguistics, which (knowing Nathan) will mean that computational analysis is a big part of the course.
,
I just wanted to let everyone know that we have an effort underway to have a conversation with members of the (NA)ACL leadership, with the goal of coming to the Fall 2017 conference at UMass with a concrete proposal for how co-location could work.
I think efforts like this are very much called for, so thanks for taking the initiative. Looking forward to seeing what emerges!
For what it’s worth, I would vote for a broadly scoped name like the above-suggested CALQ or Computational Approaches to Linguistics. Words like “Theoretical”, “Mathematical”, and “Modeling” could be taken to exclude certain computational perspectives/approaches.
To chime in to Nathan’s point on “taken to exclude certain computational perspectives/approaches” — a possible example is computational methods for corpus linguistics, which can apply to many different areas, such as sociolinguistics (like all the work on social media as corpora). This doesn’t fit into the traditional conception of cognitive science or mathematical/theoretical linguistics, as far as I understand those research traditions.