HQ is an open-source stack for annotating data with the help of LLMs. Using schemas written in natural language, items such as news articles or CSVs can be turned into dashboards that are easy to interpret, export, and audit. The aim is to help researchers apply their analytical frameworks to larger datasets with less effort. Because schemas are defined in natural language, the approach is accessible even without programming skills, giving many people a way to extract information directly.
Whether you are doing keyword labelling, 1–10 rating scales, or graph triplets, the goal is to make it easier to conduct research, gather evidence, or simply organise data. This applies whether you are working with thousands of court documents, news articles, CSV datasets, or photos of receipts.
More information and examples:
https://github.com/open-politics/open-politics-hq
https://docs.open-politics.org
The workshop will focus on presenting existing analyses and developing new ones together. The objective is for participants to gain an understanding of the software and the analytical possibilities it enables. We are also interested in hearing your feature wish-list and, where appropriate, incorporating it into the project roadmap.
We will conclude with an overview of our long-term vision, including where the tool is headed and what the next development steps will be.
P.S. Although our project also involves extensive research on validating and de-biasing LLM-generated labels, this workshop will concentrate on the practical use of HQ. For methodological discussions about robustness, please contact engage@open-politics.org