Dong Nguyen

Date: 16:00 – 16:30, Thursday, 22.04.2021
Location: MS Teams ICS Colloquium

Title: Representations and measurements in NLP
Abstract: The way language is represented in Natural Language Processing (NLP) systems has changed radically with the emergence of neural network approaches that learn to represent words, sentences, and other linguistic units as dense real-valued vectors. These vector representations are a core component of almost all modern NLP systems, and they are also increasingly used as research objects themselves. However, the representations encode societal biases, are hard to interpret, and do not capture the social aspects of language use well.
In this talk, I will focus on these challenges, especially in the context of NLP for measuring social phenomena. I will discuss ongoing projects in these areas, including work on biases in vector representations, the modeling and measurement of style, and visualization of vector spaces.