Elamur Research⌕
☰
home / papers / 2026 / Supercharging Agenda Setting Research: The ParlaCAP Dataset of 28 European Parliaments and a Scalable Multilingual LLM-Based Classification

Supercharging Agenda Setting Research: The ParlaCAP Dataset of 28 European Parliaments and a Scalable Multilingual LLM-Based Classification

Paper · 2026 · Taja Kuzman Pungeršek, Peter Rupnik, Daniela Širinić et al.

Summary

Comparing how European parliaments divide their attention across policy areas has been difficult. Manual coding under the Comparative Agendas Project (CAP) scheme is slow, and existing resources cover few countries and short periods. The authors address this with ParlaCAP, a resource built on the ParlaMint corpora that covers more than 8 million speeches from 28 parliaments.

They use an LLM teacher-student setup. GPT-4o labels a sample of speeches with 21 CAP major topics plus an extra "Other" category. A multilingual BERT-like encoder is then fine-tuned on these labels so the full collection can be classified cheaply. Human-annotated test sets in four languages show that GPT-4o agrees with human coders about as well as humans agree with each other. The fine-tuned student performs close to the LLM and beats existing CAP classifiers trained on manually coded data from other domains.

The released dataset adds ParlaSent sentiment scores and speaker and party metadata. Three example analyses look at topic attention, emotional tone and gender gaps. The main lesson is that in-domain LLM-labelled training data can outperform out-of-domain human-labelled data, with manual effort needed only for evaluation.

Key points

  • Training data: 1,200 speeches were drawn from each of the 29 ParlaMint 4.1 corpora and labelled by GPT-4o through the OpenAI batch API for about $100.
  • Final training set: 35,579 speeches, split into 29,779 for training and 5,800 for development.
  • Public Lands augmentation: this rare label was boosted by keyword search in ParlaMint.en-ana followed by GPT-4o filtering, raising its examples from 145 to 924.
  • Evaluation: manually annotated test sets ParlaCAP-test-EN, -HR, -SR and -BS were used, each with 22 labels; the EN and HR sets contain roughly 870 speeches each.
  • Releases: the classifier is on Hugging Face and the dataset is in the CROSSDA repository. The dataset is enriched with PartyFacts and V-DEM metadata.
  • Gender use case: women speak relatively more on Health, Social Welfare and Education, while men dominate transport, economy and defence topics.

Our summary of the paper, in our own words — not its text. The work belongs to its authors; read it at the links above.

 Fields, topics and concepts

📚 Shelves 6The field of science this work belongs to (OpenAlex fields).Computer Science · 16,667 worksArtificial Intelligence · 16,194 worksArts and Humanities · 16,039 worksLanguage and Linguistics · 15,891 worksSocial Sciences · 1,772 worksGeneral Social Sciences · 227 works

🧭 Topics 1Research subjects and languages the work is about.Multilingual & cross-lingual NLP · 828 works

Cite

Cite as Taja Kuzman Pungeršek, Peter Rupnik, Daniela Širinić and Nikola Ljubešić. 2026. Supercharging Agenda Setting Research: The ParlaCAP Dataset of 28 European Parliaments and a Scalable Multilingual LLM-Based Classification. Proceedings of the 3rd Workshop on Natural Language Processing for Political Sciences (PoliticalNLP 2026).

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.63317/4t2m5y2szdw8

@inproceedings{pungersek2026supercharging,
  title = {{Supercharging Agenda Setting Research: The ParlaCAP Dataset of 28 European Parliaments and a Scalable Multilingual LLM-Based Classification}},
  author = {Taja Kuzman Pungeršek, Peter Rupnik, Daniela Širinić and Nikola Ljubešić},
  booktitle = {Proceedings of the 3rd Workshop on Natural Language Processing for Political Sciences (PoliticalNLP 2026)},
  year = {2026},
  doi = {10.63317/4t2m5y2szdw8},
  url = {https://doi.org/10.63317/4t2m5y2szdw8}
}

References

32 references, 1 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Frank R Baumgartner, Christian Breunig, and Emiliano Grossman. 2019. Comparative Policy Agendas: Theory, Tools, Data. Oxford University Press. find
  2. Shaun Bevan. 2019. Gone fishing. Comparative policy agendas: Theory, tools, data, page 17. find
  3. Alexis Conneau, Kartikay Khandelwal, Naman Goyal, Vishrav Chaudhary, Guillaume Wenzek, Francisco Guzmán, Edouard Grave, Myle Ott, Luke Zettlemoyer, and Veselin Stoyanov. 2020. Unsupervised cross-lingual representation learning at scale. In Proceedings of the 58th annual meeting of the association fo… find
  4. Michael Coppedge, John Gerring, Carl Henrik Knutsen, Staffan I. Lindberg, Jan Teorell, David Altman, Fabio Angiolillo, Michael Bernhard, Agnes Cornell, M. Steven Fish, Linnea Fox, Lisa Gastaldi, Haakon Gjerløw, Adam Glynn, Ana Good God, Sandra Grahn, Allen Hicken, Katrin Kinzelbach, Kyle L. Marquard… find
  5. Franciska de Jong, Dieter Van Uytvanck, Francesca Frontini, Antal van den Bosch, Darja Fišer, and Andreas Witt. 2022. Language Matters: The European Research Infrastructure CLARIN, Today and Tomorrow. In CLARIN, pages 31–58. De Gruyter. find
  6. Benjamin de Vet and Robin Devroe. 2023. Party Control, Intraparty Competition, and the Substantive Focus of Women’s Parliamentary Questions: Evidence from Belgium. Politics and Gender, 19(1):247–271. find
  7. Holger Döring and Sven Regel. 2019. Party Facts: A database of political parties worldwide. Party Politics, 25(2):97–109. find
  8. Tomaž Erjavec, Matyáš Kopp, Nikola Ljubešić, Taja Kuzman, Paul Rayson, Petya Osenova, Maciej Ogrodniczuk, Çağrı Çöltekin, Danijel Koržinek, Katja Meden, Jure Skubic, Peter Rupnik, Tommaso Agnoloni, José Aires, Barkarson, StarkaDur, Roberto Bartolini, Núria Bel, ‘ Calzada María Pérez, Roberts Dar g… find
  9. Tomaž Erjavec, Maciej Ogrodniczuk, Petya Osenova, Nikola Ljubešić, Kiril Simov, Andrej Pančur, Michał Rudolf, Matyáš Kopp, StarkaDur Barkarson, Steinþór Steingrímsson, Çağrı Çöltekin, Jesse de Does, Katrien Depuydt, Tommaso Agnoloni, Giulia Venturi, María Calzada Pérez, Luciana D. de Macedo, Cost… find
  10. Will Jennings, Shaun Bevan, Arco Timmear-mans, Gerard Breeman, Sylvain Brouard, Laura Chaqués-Bonafont, Christoffer Green-Pedersen, Peter John, Peter B. Mortensen, and Anna M. Palau. 2011. Effects of the core functions of government on the diversity of executive agendas. Comparative Political Studie… find
  11. Klaus Krippendorff. 2018. Content analysis: An introduction to its methodology. Sage Publications. find
  12. Taja Kuzman and Nikola Ljubešić. 2025. LLM Teacher-Student Framework for Text Classification With No Manually Annotated Data: A Case Study in IPTC News Topic Classification. IEEE Access. find
  13. Taja Kuzman Pungeršek, Peter Rupnik, Ivan Porupski, Vuk Dinić, and Nikola Ljubešić. 2025. State of the Art in Text Classification for South Slavic Languages: Fine-Tuning or Prompting? arXiv preprint arXiv:2511.07989.
  14. Michal Mochtak and Maurits Meijers. In review. Finding the Needle in a Haystack: Using Large Language Models to Detect Rare Speech Acts. Journal of Computational Social Science. Michal Mochtak, Peter Rupnik, and Nikola Ljubešić. 2024. The ParlaSent Multilingual Training Dataset for Sentiment Identi… find
  15. Costanza Navarretta, Dorte Haltrup Hansen, and Bart Jongejan. 2024. Enriching the ParlaMint-DK corpus with Policy Domains. In CLARIN Annual Conference Proceedings, page 130. find
  16. OpenAI. 2024. Hello GPT-4o. https://openai.com/index/hello-gpt-4o/. Accessed September 11, 2024. ↗
  17. Christian Rauh and Jan Schwalbach. 2020. The ParlSpeech V2 data set: Full-text corpora of 6.3 million parliamentary speeches in the key legislative chambers of nine representative democracies. Harvard Dataverse, 1(1). find
  18. Christopher Michael Rytting, Taylor Sorensen, Lisa Argyle, Ethan Busby, Nancy Fulda, Joshua Gubler, and David Wingate. 2023. Towards Coding Social Science Datasets with Language Models. arXiv preprint arXiv:2306.02177. — [arXiv:2306.02177](https://arxiv.org/abs/2306.02177) find
  19. Miklós Sebők, Ákos Máté, Orsolya Ring, Viktor Kovács, and Richárd Lehoczki. 2024. Leveraging Open Large Language Models for Multilingual Policy Topic Classification: The Babel Machine Approach. Social Science Computer Review. find
  20. Miklós Sebők, Sven-Oliver Proksch, Christian Rauh, Péter Visnovitz, Gergő Balázs, and Jan Schwalbach. 2023. Comparative European Legislative Research In The Age Of Large-Scale Computational Text Analysis: A Review Article. International Political Science Review. find
  21. Daniela Širinić and Dario Nikić Čakar. 2019. Croatian political agendas. In Comparative Policy Agendas: Theory, Tools, Data. Oxford University Press. find
  22. Željko Poljak. 2024. Give the media what they need: Negativity as a media access tool for politicians. The International Journal of Press/Politics, 0(0):19401612241234861. 8. find
  23. Language Resource References Erjavec, Tomaž and Kopp, Matyáš and Kuzman Pungeršek, Taja and Ljubešić, Nikola and Ogrodniczuk, Maciej and Osenova, Petya and ... and Fišer, Darja. 2025. Multilingual comparable corpora of parliamentary debates ParlaMint 5.0. Slovenian language resource repository CLAR… find
  24. Erjavec, Tomaž and Kopp, Matyáš and Ogrodniczuk, Maciej and Osenova, Petya and ... and Fišer, Darja. 2024. Multilingual comparable corpora of parliamentary debates ParlaMint 4.1. Slovenian language resource repository CLARIN.SI. PID http://hdl.handle.net/11356/1912. ↗
  25. Erjavec, Tomaž and Kopp, Matyáš and Ogrodniczuk, Maciej and Osenova, Petya and Fišer, Darja and ... and Kryvenko, Anna. 2023. Multilingual comparable corpora of parliamentary debates ParlaMint 3.0. Slovenian language resource repository CLARIN.SI. PID http://hdl.handle.net/11356/1486. ↗
  26. Koehn, Philipp. 2011. European parliament proceedings parallel corpus 1996-2011. find
  27. Kuzman, Taja and Ljubešić, Nikola and Erjavec, Tomaž and Kopp, Matyáš and Ogrodniczuk, Maciej and Osenova, Petya and Rayson, Paul and ... and Fišer, Darja. 2024. Linguistically annotated multilingual comparable corpora of parliamentary debates in English ParlaMint-en.ana 4.1. Slovenian language res… find
  28. Kuzman Pungeršek, Taja and Ljubešić, Nikola. 2025. Multilingual ParlaCAP model for CAP Topic Classification in Parliamentary Speeches. Hugging Face. PID https://doi.org/10.57967/hf/6684. — [doi:10.57967/hf/6684](https://doi.org/10.57967/hf/6684) ↗
  29. Kuzman Pungeršek, Taja and Ljubešić, Nikola. 2026. Multilingual training dataset for CAP policy topic classification ParlaCAP-train. Slovenian language resource repository CLARIN.SI. PID http://hdl.handle.net/11356/2093. ↗
  30. Ljubešić, Nikola and Rupnik, Peter and Kuzman Pungeršek, Taja and Porupski, Ivan and Mochtak, Michal and Dinić, Vuk and Širinić, Daniela and Kopp, Matyáš and Erjavec, Tomaž. 2025. ParlaCAP: Dataset for tracking political agenda-setting across European parliaments. CROSSDA. PID https://doi.org/10.… — [doi:10.23669/1ztelp](https://doi.org/10.23669/1ztelp) find
  31. Ljubešić, Nikola and Rupnik, Peter and van Noord, Rik. 2023. Multilingual parliamentary model XLM-R-parla. Hugging Face. PID https://doi.org/10.57967/hf/6717. — [doi:10.57967/hf/6717](https://doi.org/10.57967/hf/6717) ↗
  32. Rupnik, Peter and Ljubešić, Nikola and Mochtak, Michal. 2023. Multilingual parliament sentiment regression model XLM-R-ParlaSent. Hugging Face. PID https://doi.org/10.57967/hf/6718.9. — [doi:10.57967/hf/6718.9](https://doi.org/10.57967/hf/6718.9) ↗

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/pungersek-2026-supercharging-agenda-setting-research-parlacap/