Elamur Research⌕
☰
home / papers / 2026 / SouDeC: Source Detection and Classification in Czech

SouDeC: Source Detection and Classification in Czech

Summary

Trustworthy journalism depends on knowing who stands behind reported claims, but Czech had no ready-made tool for finding and categorising such attribution sources automatically. SouDeC fills this gap. It takes plain text, has UDPipe 2 build Universal Dependencies trees, and then searches those trees for attribution signals. The signals come from a list of about 200 lemmas derived from the SiR 1.0 corpus of iRozhlas news articles. Many entries carry hand-written syntactic and contextual constraints. Each entry also has a reliability score based on training counts, which works as an adjustable threshold. Once a signal is matched, the system finds its source by walking the dependency tree. Named entities from NameTag 3 then help assign each source to one of five classes, from anonymous to official political. The authors report detection quality comparable to systems for other languages, though the exact scores are not in the supplied text. They see full coreference resolution as the main way to improve classification. A case study checks whether attribution patterns can separate low-reputation texts from reputable news.

Key points

  • Data: SiR 1.0 contains 1,718 Czech news articles (42,890 sentences, 614,995 words) with crowdsourced annotations of linked signals and sources.
  • Test sets: Development and evaluation data come from 46 triple-annotated articles plus 80 manually corrected double-annotated articles, split evenly between the two sets.
  • Source taxonomy: Sources fall into five classes: fully anonymous, only partly anonymous, unofficial, and two official groups, one non-political and one political.
  • Case study: Chain mails differed clearly from the full archive of a reputable news outlet. They could not be distinguished from that outlet's "Commentaries" section.
  • Availability: SouDeC runs as a command-line program, a web service and a REST API server, and its code is on GitHub (ufal/soudec) under the Mozilla Public License 2.

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: Linguistics & languages

Cite

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.63317/3xvp6edtnr5g

@inproceedings{mirovsku2026soudec,
  title = {{SouDeC: Source Detection and Classification in Czech}},
  author = {Mírovský, Jiří and Hladka, Barbora},
  booktitle = {Proceedings of the Fifteenth Language Resources and Evaluation Conference},
  year = {2026},
  doi = {10.63317/3xvp6edtnr5g},
  url = {https://doi.org/10.63317/3xvp6edtnr5g}
}

References

14 references, 2 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Mariana S. C. Almeida, Miguel B. Almeida, and André F. T. Martins. 2014. A joint model for quotation attribution and coreference resolution. In Proceedings of the 14th Conference of the European Chapter of the Association for Computational Linguistics, pages 39–48, Gothenburg, Sweden. Association fo… find
  2. Pouliquen Bruno, Steinberger Ralf, and Best Clive. 2007. Automatic detection of quotations in multilingual news. find
  3. Marie-Catherine De Marneffe, Christopher D Manning, Joakim Nivre, and Daniel Zeman. 2021. Universal dependencies. Computational linguistics, 47(2):255–308. find our card
  4. Barbora Hladká, Jiří Mírovský, Matyáš Kopp, and Václav Moravec. 2022. Annotating attribution in Czech news server articles. In Proceedings of the 13th Conference on Language Resources and Evaluation (LREC 2022), pages 1817–1823, Marseille, France. European Language Resources Association. find
  5. Maciej Janicki, Antti Kanner, and Eetu Mäkelä. 2023. Detection and attribution of quotes in finnish news media: Bert vs. rule-based approach. In Proceedings of the 24th Nordic Conference on Computational Linguistics (NoDaLiDa), number 52 in NEALT Proceedings Series, pages 52–59, Estonia. University… find
  6. Timothy O’Keefe, Silvia Pareti, James R. Curran, Irena Koprinska, and Matthew Honnibal. 2012. A sequence labelling approach to quote attribution. In Proceedings of the 2012 Joint Conference on Empirical Methods in Natural Language Processing and Computational Natural Language Learning, pages 790–799… find
  7. Silvia Pareti. 2012. A database of attribution relations. In Proceedings of the Eighth International Conference on Language Resources and Evaluation (LREC’12), pages 3213–3217, Istanbul, Turkey. European Language Resources Association (ELRA). find
  8. Silvia Pareti, Tim O’Keefe, Ioannis Konstas, James R. Curran, and Irena Koprinska. 2013. Automatically detecting and attributing indirect quotations. In Proceedings of the 2013 Conference on Empirical Methods in Natural Language Processing, pages 989–999, Seattle, Washington, USA. Association for Co… find
  9. Fynn Petersen-Frey and Chris Biemann. 2024. Fine-grained quotation detection and attribution in German news articles. In Proceedings of the 20th Conference on Natural Language Processing (KONVENS 2024), pages 196–208, Vienna, Austria. Association for Computational Linguistics. find
  10. Lucie Poláková, Pavlína Jínová, and Jiří Mírovský. 2015. Signals of attribution in the prague dependency treebank. In 14th International Workshop on Treebanks and Linguistic Theories (TLT 2015), pages 292–299, Warszawa, Poland. IPI-PAN, IPIPAN. find
  11. Milan Straka. 2018. UDPipe 2.0 prototype at CoNLL 2018 UD shared task. In Proceedings of the CoNLL 2018 Shared Task: Multilingual Parsing from Raw Text to Universal Dependencies, pages 197–207, Brussels, Belgium. Association for Computational Linguistics. find our card
  12. Jana Straková, Milan Straka, and Jan Hajič. 2019. Neural architectures for nested ner through linearization. In Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics, pages 5326–5331, Stroudsburg, PA, USA. Association for Computational Linguistics. 8. find
  13. Language Resource References Eduard Bejček and Eva Hajičová and Jan Hajič and Pavlína Jínová and Václava Kettnerová and Veronika Kolářová and Marie Mikulová and Jiří Mírovský and Anna Nedoluzhko and Jarmila Panevová and Lucie Poláková and Magda Ševčíková and Jan Štěpánek and Šárka Zikánová. 2013. Pr… find
  14. Barbora Hladká and Jiří Mírovský and Matyáš Kopp and Václav Moravec. 2022. SiR 1.0. Institute of Formal and Applied Linguistics, Charles University, Prague, Czech Republic, Lindat/CLARIAH, http://hdl.handle.net/11234/1-4840. ↗

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/mirovsky-2026-soudec-source-detection-classification-czech/