Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 2: Short Papers)
Why this work is in the frame
A frame that forgets how it found something cannot be audited. These are the routes that admitted this work.
Bibliographic record
Abstract
Message from the Program ChairsIt's hard to believe that we're actually going to be seeing the program come together in Toronto.We're really looking forward to it and to seeing you all there!Most of the work of a program chair is behind the scenes: herding reviewers and chairs, wrangling data from various sources, and answering lots and lots of email.This is a volunteer position, so the only reward we get for this is our chance to make the process of submitting and reviewing papers to our conference better.This letter will outline some of those experiments.First, we asked reviewers for two scores: soundness and excitement.Our goal was that any sound paper would be accepted to some ACL affiliated venue, but that the "main conference" distinction (limited by space) would be focused on the most exciting papers.Our hope was that soundness would be less noisy than a single "overall recommendation" score, which would help reduce the randomness of decisions.Judging by the exit surveys, this change was well received: over 80% of the chairs, reviewers and authors either expressed support or did not object to this change.Next, we developed a new process for matching papers to reviewers based on keywords for not only the subject matter of the paper, but also its type of contribution and target language(s).This allowed more fine-grained control over the paper-reviewer matches, and we were also able to provide the chairs with context for the paper-reviewer matches.To improve review quality, we also updated the reviewer guidelines, and developed a system for the authors to flag specific types of issues with reviews.Finally, we have also proposed a new initiative for recognizing outstanding reviewers and chairs (73 awards at ACL'23).Finally, we have tried to give more options for presentations.Findings papers now have an in-person presentation spotlight slot and virtual posters in addition to recording videos.Virtual posters have portals to link in-person attendees to virtual posters.We have also brought back Miniconf and RocketChat to allow for better virtual communication between papers (regardless of where the authors are).This conference is a result of the joint efforts of over ten thousand people.We deeply thank them all, and apologize for the many nagging emails we had to send out.In particular: AreasTo ensure a smooth process, the submissions to ACL 2023 were divided into 26 areas.The areas mostly followed these of previous ACL, and more broadly *ACL conferences, reflecting the typical divisions in the field.Following EMNLP 2022, we split the "Large Language Models" track away from "Machine learning in NLP", reflecting the growth of submissions in the area.We also offered two new tracks ("Linguistic diversity" and "Multilingualism and Cross-Lingual NLP").For the papers authored by SACs, the final recommendation decisions were made by a separate SAC team.The most popular areas (with over 250 submissions) were "Dialogue and Interactive Systems", "Information Extraction", "Large Language Models", "Machine Learning for NLP", and "NLP Applications". Best Paper AwardsACL'23 implemented the new ACL award policy, aiming to expand the pool of work that is recognized as outstanding.In total, 73 papers were nominated by the reviewers or area chairs for consideration for awards.These papers were assessed by the Best Paper Award Committee, and with their help we selected 4 best papers, 3 special awards (social impact, resource, reproduction), and several dozen outstanding papers.The best and outstanding papers will be announced in a dedicated plenary session for Best Paper Awards on July 10 2023. Presentation ModeIn ACL 2023, there is no meaningful distinction between oral and poster presentations in terms of paper quality.The composition of the oral sessions were proposed by the SACs of their respective tracks, so as to compose a thematically coherent set of papers on a shared topic or method, which would allow for an engaging discussion.The decisions were not based on the authors' virtual or on-site attendance.We hope you enjoy the program and the new elements we introduced (but let us know either way).We are looking forward to a great ACL 2023!
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
Full frame distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.001 | 0.003 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.000 | 0.001 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.000 | 0.000 |
| Open science | 0.002 | 0.001 |
| Research integrity | 0.000 | 0.000 |
| Insufficient payload (model declined to judge) | 0.000 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from it