A Bird's Eye View of EMNLP 2021

This page showcases blogposts by participants of the PhD course "Communicating State-of-the-art NLP Research to a Broader Audience" held at ITU in November 2021:

Even with videos on 2x, ingesting all the findings from the recent Conference on Empirical Methods for Natural Language Processing (EMNLP) was a daunting task. Through a synchronous workshop with PhDs from across Danish (and some European) universities, we tried to identify larger trends from different subdisciplines and distill them for a broader audience under the guidance of our invited scientific communicator Carolin Lawrence.

Depth and Breadth

Thanks to everyone's hard work, we can now present you with a collection of highlighted blog posts on a wide array of research from EMNLP:

Efficiency@EMNLP2021 for Everyone
Andreas Holm

About Biases in the Data and How That Affects The Factual Knowledge Language Models Learn
Constanza Fierro

I Agree with "Agreeing to Disagree"
David Sasu

How Robust Are Transformers?
Dennis Ulmer

Just Act Natural
Max Müller-Eberstein

A Glance at the Uniform Information Density Hypothesis in Language Processing and Speech Production
Nina Schneidermann

A Summarization of Summarization Papers from EMNLP 2021
Sotaro Takeshita

Crowdsourcing Annotations for Subjective NLP Tasks: Disagreement Is Signal, Not Noise
Terne Sasha Thorn Jakobsen

Generating or Ensembling Adapters for Low-resource Languages
Tommaso Green

We encourage you to dive deep into each of the topics above, since they helped us come out of the conference with a broader picture of the field than from just sticking to our usual tracks. Once again, thank you to all the participants for the great discussions!