Minnesota Natural Language Processing Seminar Series: Understanding Styles in NLP

The Minnesota Natural Language Processing (NLP) Seminar is a venue for faculty, postdocs, students, and anyone else interested in theoretical, computational, and human-centric aspects of natural language processing to exchange ideas and foster collaboration. The talks are every other Friday from 12 p.m. - 1 p.m. during the Fall 2021 semester.

This week's speaker, Shirley Anugrah Hayati (Georgia Institute of Technology), will be giving a talk titled "Understanding Styles in NLP."

Abstract

People use various styles, word choices, and communication strategies to express themselves more effectively. Huge pretrained language models, such as BERT, have been well known to be successful in reaching high performance for predicting these linguistic styles (e.g., politeness, formality, emotions). However, do these models learn stylistic cues as humans perceive? To answer this question, we develop a new dataset of human perception on top of benchmarking datasets of linguistics styles. Then, I will talk about how we explore whether these perception scores can provide better explanation on the models. I will conclude with a reformulation of style and its components and how we envision the future direction of style analysis in NLP.

Biography

Shirley Anugrah Hayati is a Ph.D. student in human-centered computing at the School of Interactive Computing, Georgia Institute of Technology. She received her M.S. in language technologies from Carnegie Mellon University and B.S. in computer science from Universitas Indonesia. Her research interests include human-centered NLP and computational social science. She has been a reviewer for multiple ACL conferences, actively promotes diversity in computer science by participating in CMU Women@SCS, Grace Hopper celebration, Widening NLP, and currently serves as a diversity & inclusion chair for NAACL 2022. Her current work focuses on multi-style analysis and NLP for social good.