PolBERT: A Suite of Pre-Trained Large Language Models for Political Science Research

• In Event: Large Language Model for Political Texts and Concepts

Brief Overview

We introduce PolBERT, a set of transformer models pre-trained on political text corpora, and show how domain-specific pre-training improves classification performance.

Authors

• Mitchell Bosley, University of Michigan

• Musashi Jacobs-Harukawa, Princeton University

• Alexander Hoyle, University of Maryland

• Hauke Licht, University of Cologne

• Patrick Y. Wu, New York University