At the workshop, Kirill Solovev from the University of Graz presented the pipeline developed in VALPOP to extract political elite networks from news articles, with a pilot study conducted for Poland. The workshop was held as part of sessions organised by the European Consortium for Political Research (ECPR).

Read the paper abstract: Mapping Political-Elite Networks in Europe with a Multilingual Joint Entity-Relation Extraction Pipeline: The VALPOP Project.

Extracting people and their relations from news articles

The pipeline is a flexible system designed to extract political relationships from multilingual text. It identifies people and organisations and maps the connections between them, while allowing researchers to modify individual components without compromising the consistency and comparability of results. In its initial phase, the system has been tested on Polish news data, extracting over one million relationships from approximately 250,000 articles. The result exposed well-known features of Polish politics, such as the close ties between government and state companies and the strong rivalry between the two main parties, Civic Platform and Law and Justice.

Defining the use of the pipeline

The next step is to derive the network measures that VALPOP will use to compare countries, including centrality (who are the most influential actors), density (how tightly connected is the network), and the classification of networks (is the network closed or open) along the O-group/P-group distinction (distributional coalitions versus civic associations), a typology operationalised for rule-of-law research by our sister Horizon project NetROL. Check out their working paper: Networks and the Rule of Law: Conceptual Framework.

In total, nearly 180 million public articles from media, business, and government sources are expected to feed into the datasets. These datasets will underpin the exploration of how political elite networks relate to levels of populism, and how these dynamics connect to national variations in the rule of law, particularly in the valuation and distribution of public goods.

Ultimately, the project will deliver a text2network tool that systematically and automatically identifies and extracts network relationships from unstructured text.