ML@GT Further Establishes Itself in Natural Language Processing Community
Over the years, the Machine Learning Center at Georgia Tech (ML@GT) has steadily been increasing its presence in the natural language processing (NLP) community. With a few new hires this fall and a big performance at this year’s Conference on Empirical Methods in Natural Language Processing (EMNLP), ML@GT is becoming a major player in the field.
With 10 papers accepted to the main EMNLP conference and three papers at a new, coinciding sister publication called Findings of EMNLP, Georgia Tech researchers are tackling problems like how to summarize text and interpreting how persuasive language impacts domains advertising and argumentation.
Diyi Yang, an assistant professor at ML@GT and the School of Interactive Computing (IC) leads the Georgia Tech group with six accepted papers. Yang is also the area chair for the interpretability and analysis of models for the NLP track.
“EMNLP is one of the most prestigious conferences in NLP, and one of the best conferences for the interaction of machine learning and NLP. Georgia Tech’s NLP program is emerging and our presence at EMNLP is a strong signal of it,” said Yang.
Yang also noted that many of the papers accepted from her lab, Social Language and Technologies (SALT), are from first-time authors who are also first-year Ph.D., master’s, or undergraduate students.
“It is a great recognition of our students’ passion, efforts, and their research impact in this NLP space. I feel proud of them and excited about more high-impact and fun projects that will come next!” said Yang.
Outside of accepted papers, Yang is also an invited speaker for the 4th Workshop on NLP and CSS. ML@GT Ph.D. student Yuval Pinter is co-organizing BlackboxNLP, a workshop at EMNLP dedicated to model interpretability. IC Associate Professor Alan Ritter and IC Assistant Professor Wei “Coco” Xu are co-organizing the Workshop on Noisy Text. Ritter and Wei are also involved with the language generation track as a senior area chair for information extraction and area chair, respectively.
Overall, EMNLP accepted 602 long papers and 151 short papers, with a 22.4 percent and 15.5 percent acceptance rate respectively. The conference will take place online Nov. 16 through 20.