Welcome to APSI — the Automated Political Stance Identification Tool.
APSI provides a theory-driven and transparent approach to identify political positions in texts. The tool uses pre-trained language models to estimate where texts fall along key political dimensions, including economic ideology, support for liberal democratic values, and populist versus pluralist rhetoric. APSI focuses on linguistic patterns and provides probabilistic scores. Our goal is to support research, journalism, non-governmental organization and anyone interested in the critical exploration of political discourse.
What the tool does
The Automated Political Stance Identifier (APSI) is a tool designed to analyse political texts and identify the stances they express. It uses a hypothesis-based approach grounded in political science theories to evaluate whether a text reflects specific political stances, such as economic left–right orientation, populist versus pluralist rhetoric, and support for liberal democratic values. Instead of relying on manually labeled training data, APSI applies a Natural Language Inference (NLI) model to test how strongly a text supports or contradicts a set of predefined hypotheses. By aggregating these results, the tool produces interpretable stance scores, confidence measures, and supporting evidence, enabling researchers, policymakers, and the public to systematically evaluate political discourse in a transparent and reproducible way.
The dimensions we measure
Economic Left–Right Scale
It captures positions ranging from strong state intervention, redistribution, and welfare provision to free markets, privatization, and a limited role for government.
Support for Liberal Democracy
It captures the level of commitment to free and fair elections, political competition, and fundamental rights such as freedom of speech, media, assembly, and association expressed in the text.
Populism–Pluralism Rhetoric
It captures whether the text emphasizes the will of “the people” against elites and institutions, or instead supports institutional constraints, and minority rights.
How the analysis works
The APSI tool uses a theory-driven Natural Language Inference (NLI) approach to analyse political texts. Instead of relying on keywords or labeled training data, the tool evaluates how strongly a text supports or contradicts predefined political statements in the form of hypotheses derived from political science theories. It first checks whether a text is relevant, then estimates its position across key dimensions — such as economic ideology, populist versus pluralist rhetoric, and support for liberal democratic values. These results are combined into transparent scores, confidence estimates, and supporting evidence.
Ethical considerations
This tool is for research and information only. Like any AI system, it can be biased due to its training data and how it is designed.
It should not be used for profiling, surveillance, or making decisions about people.
