Semantic Scaling: Bayesian Ideal Point Estimates with Large Language Models

arXiv:2405.02472v1 Announce Type: new
Abstract: This paper introduces "Semantic Scaling," a novel method for ideal point estimation from text. I leverage large language models to classify documents based on their expressed stances and extract survey-like data. I then use item response theory to scale subjects from these data. Semantic Scaling significantly improves on existing text-based scaling methods, and allows researchers to explicitly define the ideological dimensions they measure. This represents the first scaling approach that allows such flexibility outside of …

abstract arxiv bayesian cs.cl data documents extract language language models large language large language models novel paper scale scaling semantic survey text theory type