Measuring Policy Positions from Text

Guest Lecture — University of Bologna, PhD Cycle 41

Author

Paride Carrara

Published

June 7, 2026

1 About this lecture

This notebook accompanies a guest lecture delivered at the University of Bologna for the PhD Programme in Political and Social Sciences on measuring policy positions from political text, using Italian parliamentary speeches as a running example.

The lecture walks through five methods, from simple Bag of Words approaches to LLMs:

Module Topic
Pre-requisites R packages, Python, and OLLAMA setup
Module 1 Text as data & bag-of-words
Module 2 Wordscores & Wordfish
Module 3 Latent Semantic Scaling
Module 4 Generative LLMs via OLLAMA
Module 5 BERT models (ManifestoBerta)

2 How to use

Each chapter is also a standalone Quarto notebook. You can render any single module on its own with quarto render 0X_<name>.qmd, or build the entire book with quarto render.

The slide decks shown during the lecture live separately in the slides/ folder of the source repository.

3 Author

Paride Carrara — political scientist currently post-Doctoral researcher at the University of Latvia for the WICE project. Research interests include party politics, voter-behaviour and computational methods.