Multilingual estimation of political-party positioning: From label aggregation to long-input Transformers

TL;DR

This paper compares label aggregation and long-input Transformer models for multilingual political scale estimation from party manifestos, finding that label aggregation yields the best results across diverse languages and countries.

Contribution

It introduces and evaluates two approaches for multilingual political scaling analysis, demonstrating the effectiveness of label aggregation and Transformer models on a large, multilingual dataset.

Findings

Label aggregation outperforms Transformer models in accuracy.

State-of-the-art models handle long texts effectively.

Analysis covers 41 countries and 27 languages.

Abstract

Scaling analysis is a technique in computational political science that assigns a political actor (e.g. politician or party) a score on a predefined scale based on a (typically long) body of text (e.g. a parliamentary speech or an election manifesto). For example, political scientists have often used the left--right scale to systematically analyse political landscapes of different countries. NLP methods for automatic scaling analysis can find broad application provided they (i) are able to deal with long texts and (ii) work robustly across domains and languages. In this work, we implement and compare two approaches to automatic scaling analysis of political-party manifestos: label aggregation, a pipeline strategy relying on annotations of individual statements from the manifestos, and long-input-Transformer-based models, which compute scaling values directly from raw text. We carry out the analysis of the Comparative Manifestos Project dataset across 41 countries and 27 languages and find that the task can be efficiently solved by state-of-the-art models, with label aggregation producing the best results.