B1366
Title: The context: Determining sentiment using large language models
Authors: Christian Breitung - Technical University of Munich (Germany) [presenting]
Garvin Kruthof - Technical University of Munich (Germany)
Sebastian Mueller - Technical University of Munich (Germany)
Abstract: Traditional sentiment classification methods lack the ability to contextualize the sentiment of macroeconomic news. We address this issue and show how large language models (LLMs) may be used to assign industry-specific sentiments to macroeconomic news. We find that the contextualization ability of a language model is positively correlated with its size and varies across industries.
