[NLP for Social Science #3]: Lucy Li - Incite at Columbia University

This free public lecture series explores how NLP is used to illuminate social processes and institutions.

#3: Context-Dependent Depictions of People Across Three Domains

A common paradigm for identifying semantic differences across social and temporal contexts is the use of static word embeddings and their distances. In particular, past work has compared embeddings against "semantic axes" that represent two opposing concepts.

We extend this paradigm to BERT embeddings and construct contextualized axes that mitigate the pitfall where antonyms have neighboring representations. We validate and demonstrate these axes on multiple people-centric datasets: occupations from Wikipedia, and multi-platform discussions in extremist, men's communities over fourteen years. In both studies, contextualized semantic axes can characterize differences among instances of the same word type. In the latter study, we show that references to women and the contexts around them have become more detestable over time.

About this workshop

The goal of this new lecture series is to invite scholars whose research uses contemporary NLP methods to illuminate not only the internal structures and patterns of linguistic forms but also the social processes and institutions in which they are embedded. The work of these speakers represents some of the most promising directions in applied NLP due to its technical rigor, research design, and theoretical depth.

Support for this series is generously provided by INCITE at Columbia and the Platial Analysis Lab at McGill. This workshop is jointly organized by Jack LaViolette (Columbia University, sociology) and Mikael Brunila (McGill University, geography).

You can find the complete schedule of dates and speakers on our website.