Political career benchmark presented at PolMeth Europe 2026 in Dublin
Summary
Bastián González-Bustamante, together with Tom Bellens, Anne Bellon, Levent Demirelli, Jan-Hinrik Meyer-Sahling, and Fanni Toth, presented the paper From Text to Trajectories: Evaluating NLP Tools and LLMs for Automated Political Career Data Collection at PolMeth Europe 2026 in Dublin on 14-15 May 2026.
The paper introduces a benchmark framework for the automated reconstruction of political career trajectories and focuses on three core subtasks: data collection, data extraction, and data labelling. More broadly, it addresses a central methodological challenge in computational political science: How to evaluate NLP tools and LLMs for career data collection in a systematic, reliable, and reproducible way.
This work forms part of the CoREx COST Action and is specifically connected to Working Group 2: Career Patterns in the Executive Triangle, which studies executive politicians, advisers, and top civil bureaucrats from a comparative perspective.
Key highlight
A central highlight of the presentation was the use of the CoREx Gold Standard to evaluate the performance of both BERT models and LLMs on extraction and classification tasks. The benchmark covers ministers, state secretaries, chiefs of ministerial cabinets, managing advisers, and senior bureaucrats across Europe and beyond, comprising more than 7,000 individuals across 35 countries.
This provides a rare comparative basis for assessing how well different model families perform when transforming unstructured biographical material into structured political career data. By grounding evaluation in a manually compiled gold-standard corpus, the paper helps establish a more transparent and reproducible benchmark for future work on elite trajectories and executive career research.
