Reading a parliament as a corpus.

CLARIN 2026

PosterPhD student sessionBrighton, 30 September 2026

A multi-scale, traceable analysis of Slovenian parliamentary discourse, 1992–2022

Every claim in this thesis can be walked back to the speeches it rests on, on CLARIN.SI infrastructure, end to end.

• Poster↓ The A0 sheet, as a PDF · 0.7 MB
• Abstract↓ Two pages, with references, as a PDF · 65 KB

1. RQ1 · Macro What the Assembly discusses 213 372 speeches → 398 topics
2. RQ2 · Meso How the contested past is mobilised, traced term by term 398 topics → one term → 2 059 lines
3. RQ3 · Micro How accountability is enacted under pressure Interpellations → feature rates → extremes read in full

Read the abstractAbout 500 words

This PhD project presents a corpus read at three levels, in a chain in which every step is saved and re-runnable; topic → term list → CQL query → concordance lines → close-read speech. The approach is from distant to close (macro → meso → micro), using a “glass box” approach.

The corpus used is siParl 4.0 (Pančur et al. 2024) from the CLARIN.SI repository, more specifically, its distribution with full linguistic annotation (Ljubešić & Dobrovoljc 2019). The research focuses only on plenary speeches, and chair procedural turns are excluded, netting a total of 213 372 speeches in eight terms and 30 years. The pipeline also relies on the CLARIN.SI concordancer, for corpus analysis, and two CLASSLA resources: the Slovenian Language Technologies Overview (for embedding model candidates) and the ParlaCAP classifier (Kuzman Pungeršek et al. 2026).

The broadest, macro scale (distant reading) employs one BERTopic model over the 213 372 speeches. The default encoder reads a median 29 % of a speech (due to truncation). Five multilingual encoders were compared, each reading whole speeches as raw text, and using a lemma POS filter for the topic labels. The chosen model was Qwen3-Embedding-0.6B (runnable on a free-tier Colab T4), and the production run had a minimum cluster size of 75, where every policy area above 1 % of speeches has at least four topics (set against ParlaCAP labels).

RQ1 asks what parliament discusses, how its attention shifts, and how small, niche parties engage relative to mainstream ones.

The meso (corpus) scale picks terms from the topic model strings and runs a frequency and collocation analysis on a specific issue emerging from RQ1, selected from domain knowledge (in our case ‘war’). Collocation parameters are set: logDice ≥ 7 strong, window ±5. From the results, specific cases are chosen for further investigation, in our case “grobišče” (‘war grave’), logDice 8.92 (topic T181). A frequency matrix is created to have a broader view of emerging patterns. CQL queries allowed to map the 2 059 plenary hits into “prompted” (1 613 on sitting days of five war-graves acts and related debates) and “unprompted” (446).

At this scale, the salience and dynamics of ideological discourse are compared to scandals (persistence vs spike and fade).

The micro scale (close reading) builds on scandals, focusing on interpellation debates, which call ministers to account but rarely end in dismissal (Pegan & Krašovec 2021). An interpellation subcorpus will be created (enriching the main corpus) and specific adversarial linguistic features will be measured (deontic modality, negation, address forms and blame pronouns) over every interpellation speech, and each feature’s extremes read in full. These speeches’ topic labels tie them back to the macro scale; topic dispersion shows whether a debate stays on its policy area.

Upon completion, corpus corrections and various user-based improvement suggestions will be submitted to the siParl authors. The embeddings, topic modeling pipeline, ParlaCAP labels back to 1990, the interpellation metadata, and other deliverables will be deposited in CLARIN.SI, with documentation for re-running it on ParlaMint corpora.

Keywords

siParl; parliamentary discourse; topic modeling; collocation; critical discourse analysis; traceability

References

• Kuzman Pungeršek, T., Rupnik, P., Širinić, D., & Ljubešić, N. (2026). Supercharging agenda setting research: The ParlaCAP dataset of 28 European parliaments and a scalable multilingual LLM-based classification. arXiv:2602.16516.
• Ljubešić, N., & Dobrovoljc, K. (2019). What does neural bring? Analysing improvements in morphosyntactic annotation and lemmatisation of Slovenian, Croatian and Serbian. In Proceedings of the 7th Workshop on Balto-Slavic Natural Language Processing (pp. 29–34). Association for Computational Linguistics. https://doi.org/10.18653/v1/W19-3704
• Pančur, A., Meden, K., Erjavec, T., Ojsteršek, M., Šorn, M., & Blaj Hribar, N. (2024). Slovenian parliamentary corpus (1990–2022) siParl 4.0. Slovenian language resource repository CLARIN.SI. http://hdl.handle.net/11356/1936
• Pegan, A., & Krašovec, A. (2021). Slovenian Parliament. PSA Parliaments briefing. https://psaparliaments.org/wp-content/uploads/2021/03/Slovenia-1.pdf

Cite the posterAPA · BibTeX

APA

Bordon, D. (2026, September 30). A multi-scale, traceable analysis of Slovenian parliamentary discourse, 1992–2022 [Poster presentation]. CLARIN Annual Conference 2026, Brighton, United Kingdom. https://davidbordon.com/academia

BibTeX

@misc{bordon2026clarin,
  author       = {Bordon, David},
  title        = {A multi-scale, traceable analysis of {S}lovenian
                  parliamentary discourse, 1992--2022},
  howpublished = {Poster, CLARIN Annual Conference 2026,
                  PhD student session, Brighton, United Kingdom},
  month        = sep,
  year         = {2026},
  url          = {https://davidbordon.com/academia}
}

Current work

Publications

Edited volumes

• 2026

  Eskevich, M., Vandeghinste, V., & Bordon, D. (Eds.). ParlaCLARIN V: Workshop on Interoperability, Multilinguality, and Multimodality in Parliamentary Corpora. LREC 2026 proceedings, Palma de Mallorca. ELRA.

  Workshop Proceedings PDF

• 2024

  ParlaCLARIN IV: Workshop on Creating, Analysing, and Increasing Accessibility of Parliamentary Corpora. LREC-COLING 2024 proceedings, Torino. ELRA.

  Workshop Proceedings PDF

Articles and chapters

• 2024

  Popič, D., Bordon, D., Franza, J., & Bezgovšek, L. Jezikovna orodja za slovenščino v Italiji: Jezikovni asistent Loris 1.1. In M. Grgič (Ed.), Slovenščina v Italiji (pp. 167–194). Založba Univerze v Ljubljani.

  doi.org/10.4312/9789612974725

• 2023

  Bordon, D. Do you Speak Neuralese? How People Comprehend the Modern Language of Machine Translation Engines – expanded article. Slovenščina 2.0, 11(1), 138–159.

  doi.org/10.4312/slo2.0.2023.1.138-159

Conference papers

• 2022

  Bordon, D. Do you Speak Neuralese? How People Comprehend the Modern Language of Machine Translation Engines. In D. Fišer & T. Erjavec (Eds.), Proceedings of the Conference on Language Technologies and Digital Humanities (pp. 286–291). Ljubljana.

  PDF

  Best student paper

Conference presentations

• 2026

  Bordon, D. A Multi-Scale, Traceable Analysis of Slovenian Parliamentary Discourse, 1992–2022 [poster]. CLARIN Annual Conference 2026, PhD student session, Brighton.

  Poster PDF Abstract PDF

• 2025

  Bordon, D. Predicting Grammatical Cases in Slovenian Varieties in Italy: A Use Case from the LORIS 1.1 Language Assistant [poster]. AIUCD 2025, University of Verona.

  Poster

  Gigliozzi Prize for best poster

• 2025

  Bordon, D. The Historical Discourse of National Trauma. Helsinki Conference on Emotions, Populism and Polarisation (HEPP5), University of Helsinki.

• 2025

  Bordon, D. Linking External Events to Discourse Shifts in Slovenian Parliamentary Debates. Corpus Linguistics 2025, Aston University, Birmingham.

• 2025

  Bordon, D. Discourse Analysis of Fringe Parties in Slovenian Parliamentary Debates. "From the extremes to the centre and back again?", University of Navarra, Pamplona.

• 2025

  Bordon, D. Identifying and Categorizing Topics in Slovenian Parliamentary Debates. New Perspectives in Humanities and Social Sciences, University of Ljubljana.

• 2024

  Bordon, D., Gorjanc, V., & Anastasi, S. Narratives of a Nation: The Discourse of National Unity and Division in the Slovene Parliament. Sociolinguistic Symposium 25, Curtin University, Perth.

Language resources

• 2023

  Popič, D., Bordon, D., & Bezgovšek, L. Loris 1.0, application and web interface. Slovenski raziskovalni inštitut (SLORI).

  jeziknaklik.it/loris

Conference reports

• 2025

  Bordon, D. Konferenca Digitalna humanistika ADHO DH 2025, Lizbona. Slovenščina 2.0, 13(1), 138–146.

  doi.org/10.4312/slo2.0.2025.1.138-146

• 2023

  Bordon, D. DHH23 – Helsinki Digital Humanities Hackathon 2023. Slovenščina 2.0, 11(2), 92–101.

  doi.org/10.4312/slo2.0.2023.2.92-101

• 2022

  Bordon, D. Konferenca Jezikovne tehnologije in digitalna humanistika 2022. Slovenščina 2.0, 10(1), 131–135.

  doi.org/10.4312/slo2.0.2022.1.131-135

Awards

• 2025 Gigliozzi Prize for best poster, AIUCD 2025, Verona
• 2023 Best Poster Design, Helsinki Digital Humanities Hackathon
• 2022 Best student paper, JT-DH 2022, Ljubljana

Chairing and organisation

• 2026 ParlaCLARIN V – chair, organiser and editor. LREC 2026, Palma de Mallorca
• 2024 ParlaCLARIN IV – chair, organiser and editor. LREC-COLING 2024, Torino
• 2024 Student Section, JT-DH 2024 – chair

Teaching and training

Curriculum vitae