Abstract
Modern natural language processing techniques can be used to extract compact and transferable representations of text and audio at multiple levels of granularity. In this talk, I will present three studies using NLP methods to model individual traits for cognitive and social science applications. The first study explores applications of both traditional ML methods and more advanced text and audio encoding techniques in language-based inference of clinical conditions. The second study focuses on developing self-supervised methods to extract transferable author representations from text. The third study explores social science applications of text modeling techniques, especially focusing on the use of language models for text-based measurement of political identity.

About the speaker
Roberta Rocca
Postdoc
Interacting Minds Centre
Aarhus University