Comment 182

"I think we have some more serious methodological issues and fundamental assumptions as a ""field""

that hinder development and growth, impacting fair reviewing. It is not just the process of reviewing,

but also the content (i.e. the object of inquiry, if any) that needs to be reflected upon / reviewed.

In face of a paradigm shift, we, as a field, need to have a conversation (instead of avoidance/denial).

Otherwise, thousands more papers wouldn't mean a thing. One concrete suggestion: start an annual edition

of ""Virtual ACL"" where reviews and handling (including meta discussions and accept/revise/reject decisions)

are all open (same for both main and workshops), also allowing feedback and questions from public. Kind

of like ICLR, but even more open in regard to meta discussions and final decision process. Everything

for this Virtual ACL main program will be handled centralized via a special ""ARR for Virtual ACL""

process --- rolling review, submit anytime in the year, but authors need to commit up front that they

are only interested in presenting at Virtual ACL (to better gauge load). Workshops at Virtual ACL will

have reviewing also fully open on ARR but the organization (e.g. reviewer matching and decision process)

will be the responsibilities of the organizers --- that said, we, as a field, need to sort out some

fundamental issues, workshops running for over 10 years in a roll (e.g. SIGMORPHON) is a sign of the

field not growing or those workshops probably should find their own independent place in the scientific

system outside of a tech conference. Workshops are not annual cult meetings funded by everyone. Workshops

are supposed to start new, lesser-known initiatives and promote them to mainstream practice. Virtual

ACL will also have registration prices that are friendlier, on par with online reg at other ML conferences

thus far (and I hope my saying this will not encourage others to exploit this opportunity to hike up

their prices). [Other personal preferences/wishes: i. can we switch to single-column format for our

papers? Looks so much cleaner. ii. no more ""words""/""sentences"" or grammar or ""classical linguistics""

for this Virtual ACL* --- how about directions proper LangTech (workshops/tracks on LaTeX, emojis, (history

of) character encodings, tutorials/collaborations with Unicode Consortium, or byte space education?),

or NLP in finer granularity (chars/bytes/stats) only, or data tracks (but not for benchmarking) or data-centric

approaches, post-processing techniques and evaluation? Kind of like data and evaluation but without

the grammar initiatives. One goal is to reconnect language with other sciences and with the rest of

computing, instead of ""special casing"" it as something ""odd"", having some more obscure distributional

character, or assuming that addressing language or language data or human perspectives requires an adversarial

standpoint towards machine learning or statistical methods. (Past generations may have done that because

of ""words"" and other philological traditions, and some may have had to start ""selling"" language

because of this ""marketing practice"" in research and science that has proliferated due to various

sociological reasons. It is at least my hope that we can be normal again and do honest good science,

and have a venue for it so we can exchange with the like-minded. (I leave the subject of the necessity

of conferences open here.)) *iii. Association of Computing and Language? Or keep name as is and change

Linguistics in one fell swoop. iv. abolish anonymity policy. Just let people post on arXiv, if they

choose to. The (sociological/academico-political) situation** is complicated enough wrt posting that

we don't need to add another barrier for authors. **There are pluses/minuses to posting that individual

authors need to consider anyway.]