/ News /

Renewed commitment in support of the Efficient NLP community

I am glad to announce that I'm renewing my commitment in support of the Efficient NLP community. I believe that inventing simple and efficient solutions is critical to increasing the participation in NLP research. This will enable more scientists to contribute to the state of the art and create sustainable progress in the long run.

First, I will be an Area Chair for the Efficient NLP Track at EMNLP 2022, held in December 2022. The Efficient NLP Track encourages submissions that focus on reducing memory requirements, improving training and inference efficiency, performing efficient model selection, and coming up with new methods to measure efficiency. I served as Area Chair for the 2021 Efficient NLP Track at EMNLP, which was a great success with over 130 submissions. I believe this track is critical to promoting research that goes beyond optimizing model accuracy.

Second, I am a co-organizer of SustaiNLP 2022—the third workshop on simple and efficient natural language processing (co-located with EMNLP 2022). SustaiNLP will provide a dedicated venue and discussion platform for researchers working on efficiency topics. The 2021 edition has received 51 submissions and has hosted a number of high profile keynotes. SustaiNLP will promote research focussing on clever methods that improve upon efficiency dimensions.

Third, as a member of ACL's Efficient NLP working group, I will contribute to the implementation of the Efficient NLP policy. The ACL executive committee has approved our policy earlier this year, which will impact important aspects of future conferences. Our recommendations include: (i) Better aligning experiments and research claims by changing the review process. (ii) Rewarding the open release of models of different sizes. (iii) Setting up efficiency tracks at conferences. I believe the Efficient NLP policy marks a significant milestone for the community, which will raise the equity in NLP research and ensure that future conferences reward scientific contributions without requiring enourmous resources.