Why EMNLP-2021 is not extending the deadline

We have received requests for extending our submission deadline due to the challenges that some of our colleagues and friends are undergoing. In consultation with the SIGDAT board, we have taken the hard decision of not extending the submission deadline. In this post, we would like to provide details about the reasons that brought us to this decision.

Although we have witnessed conferences extending submission deadlines during the past year, especially in the earlier days of the global pandemic, extending the deadline by a substantial amount of time such that it would be helpful to authors in adverse situations needs to be discussed and evaluated carefully. Conference schedules are well-planned to meet multiple external and internal constraints. The initial schedule was decided based on a number of constraints that are there mainly to help authors. For example, we had pushed our paper submission deadline well after the ACL notification date to give authors a chance to improve their papers before resubmission to EMNLP. In addition, we avoided clashes with other major related conferences, for example giving enough time between EMNLP and NeurIPS deadlines to help authors who submit to both conferences.

We nevertheless had numerous discussions as an organizing committee and in consultation with SIGDAT (and ARR where relevant) to consider possible scenarios for extension and expected outcomes:

• One-month extension: This proposal was clearly not feasible because the schedule this year was already very tight. As the time reserved for the review process and proceedings production had been shortened already, giving a long extension would lower the quality of paper review and selection.
• One-week extension: We considered this to be potentially feasible even though at the time it would have clashed with the NeurIPS deadline, which we had carefully tried to avoid. It would still require to shift the entire schedule and put additional pressure and burden on the committee, of which many members are under the same difficult circumstances as authors.
• 48-hour extension: This option has also been discussed. However, the consensus was that it would be merely symbolic and not practically meaningful to help authors in need. After the abstract submission deadline, we have reevaluated it and consulted SIGDAT again. Because additional complexities for ensuring fairness to all authors outweighed the potential benefits, in the end we decided not to grant a 48-hour extension.
• Extending ARR for EMNLP-2021 deadline to June 15th: One proposal we received was to consider papers submitted to ARR by June 15th, by authors who self-declare that they have been impacted by some crisis. However, any extension would need to be granted to all authors for fairness. Extending the ARR deadline to June 15th would have encouraged many authors to withdraw their abstracts from EMNLP and submit to ARR later. This would have overloaded ARR, and the reviews would not have been ready on time for our final decisions for EMNLP.

Although it may not seem obvious externally, conference schedules also need to balance multiple factors to ensure the quality of the paper review and selection process. For instance, time needs to be reserved for reviewer-paper matching and assignment, for reviewers discussion, for authors to write their rebuttals, and for thoughtful discussion and final recommendations amongst area chairs and senior area chairs. After the final decisions of acceptance/rejection have been made, publication chairs also need time to prepare the final proceedings before the conference date. Once the schedule is decided and announced, changing it not only breaches the “social contract” between the conference and the reviewers, (senior) area-chairs and potential authors and participants, but may adversely burden all members further, compromising the quality of the process.

One criticism and rightful question we received was why we could not extend the deadline just as done last year. We have to emphasize that each year the conference is organized by a different set of committee members, under different constraints. Although we do try our best to learn from the past experience, important decisions, especially those that are not formally defined in the ACL policies, would still need to be discussed and evaluated independently each year. For instance, the contract signed with the conference venue dictates the in-person presence at certain dates. The hybrid format we are adopting this year aims for broadening the participation, but also inevitably limits the flexibility in scheduling compared to the all virtual format. We also note that the abstract deadline this year has effectively extended the full paper submission deadline by one week. We are currently assigning submissions to reviewers, running conflict-of-interest checks, and balancing the review load of each track after we collected the necessary information, while still allowing the authors to finalize their papers in parallel.

We understand that many people in our community are currently affected by adverse events, especially our colleagues in South Asia due to the dire COVID-19 crisis and those affected by the recent Israeli-Palestinian strife. However, in our view, it is not feasible nor desirable for conferences to respond to the adverse events that continuously happen in the various parts of the world by granting extensions. The very existence of academic conferences is to encourage quality research work that can be validated by peers and shared with the community in a timely fashion. While a strict submission deadline is necessary, it unfortunately adds stress to the authors. As part of the review reform process widely discussed in the ACL community, we hope that moving towards continuous submissions with the ARR system would alleviate this issue in the future.

Organizing a major conference has been a new task for each one of us. While we cannot say that all our decisions will be right, we certainly do not take the job lightly. After carefully weighing all the arguments as explained above, in consultation with the SIGDAT board, we reached the conclusion not to postpone the EMNLP-2021 submission deadline. We hope our colleagues understand our constraints and their implications on a large number of volunteers helping make the conference possible.

Program co-chairs: Xuanjing Huang, Lucia Specia, Scott Wen-tau Yih
General chair: Marie-Francine Moens