pith. sign in

arxiv: 2606.24337 · v1 · pith:WJKYGSHGnew · submitted 2026-06-23 · 💻 cs.CL

Meet UD_Czech-PDTC: A Large and Genre-Rich Treebank in Universal Dependencies

Marie Mikulov\'a , Barbora \v{S}t\v{e}p\'ankov\'a , Daniel Zeman , Jan \v{S}t\v{e}p\'anek , Milan Straka , Jan Haji\v{c} This is my paper

Pith reviewed 2026-06-26 00:02 UTC · model grok-4.3

classification 💻 cs.CL
keywords Universal DependenciesPrague Dependency TreebankCzechtreebank conversiondependency parsingmulti-layer annotationgenre diversitynatural language processing
0
0 comments X

The pith

The Prague Dependency Treebank-Consolidated converts to Universal Dependencies to form a large, genre-rich Czech treebank.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper describes the conversion of the expanded Prague Dependency Treebank-Consolidated into Universal Dependencies. PDT-C is more than twice the size of the original PDT and spans a wider range of genres and domains. The authors detail numerous small differences in dependency topology and in the granularity of POS tags and relation labels between the two schemes. They show how these differences are resolved during conversion. The multi-layer annotations already present in PDT supply every piece of information required for standard UD trees and additional layers beyond them.

Core claim

The conversion of PDT-C to UD produces UD_Czech-PDTC, a treebank more than twice as large as prior Czech UD resources and covering more genres and domains. Although PDT and UD differ in the topology of dependency structures and the granularity of their POS and relation inventories, these differences can be overcome systematically. PDT's multi-layer annotation provides all information needed for basic UD trees plus much more.

What carries the argument

The conversion mapping that resolves differences in dependency topology and label granularity between PDT's multi-layer annotations and UD dependency trees, POS tags, and relations.

If this is right

  • UD_Czech-PDTC becomes available as one of the largest Czech resources in the Universal Dependencies collection.
  • NLP tools for Czech can be trained on a dataset that includes greater genre and domain variety than before.
  • The extra annotation layers in the original PDT can be retained alongside the UD trees for richer analysis.
  • The conversion process demonstrates a workable path from language-specific multi-layer schemes to the UD format.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The same conversion approach may apply to other richly annotated treebanks that predate UD.
  • Greater genre coverage could improve the robustness of parsers and other models when tested across text types.
  • Cross-lingual UD studies may now draw on a larger and more varied Czech component for comparison.

Load-bearing premise

The small differences in dependency structures and label granularity between PDT and UD can be resolved systematically during conversion without introducing inconsistencies or losing essential information.

What would settle it

A sample of converted sentences that, when checked against direct human UD annotation, shows systematic unrecoverable mismatches in dependency arcs or lost syntactic distinctions.

Figures

Figures reproduced from arXiv: 2606.24337 by Barbora \v{S}t\v{e}p\'ankov\'a, Daniel Zeman, Jan Haji\v{c}, Jan \v{S}t\v{e}p\'anek, Marie Mikulov\'a, Milan Straka.

Figure 1
Figure 1. Figure 1: PDT sentence representation and its conversion to UD. conversion of the Prague Dependency Treebank – Consolidated, released in 2024 (PDT-C 2.0; Ha￾jič et al., 2024), and with 3440K words is now one of the largest treebanks in UD, containing texts of various genres. In addition to the pilot Prague De￾pendency Treebank of journalistic texts (enriched version from 2006; Hajič et al., 2006, converted to the fi… view at source ↗
Figure 2
Figure 2. Figure 2: Representation of the Czech sentence (1) in UD (top) and PDT (bottom) annotation scheme. In the contribution, we discuss the major dif￾ferences between the two formalisms (theoreti￾cal foundations, approaches to syntax, morphol￾ogy, and semantics). Using specific exam￾ples from various areas of language description (part-of-speech classification, coordination, ellip￾sis, semantic-pragmatic expressions), we… view at source ↗
Figure 3
Figure 3. Figure 3: UD representation of sentence (2). are solutions some similar other not [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: PDT representation of sentence (2). The main difference is that PDT uses a multi￾layer scheme separating linguistic information, while UD integrates all (mainly morphosyntactic) information into a single graph. In the follow￾ing sections, we primarily compare the UD repre￾sentation and PDT representation at the m-layer (the basis for POS and features conversion) and the a-layer (the basis for conversion of… view at source ↗
Figure 5
Figure 5. Figure 5: Basic (top) and enhanced (bottom) UD representation of sentence (3). are-they are-not-they right or [PITH_FULL_IMAGE:figures/full_fig_p006_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: PDT representation of sentence (3). Despite the different representation of coordina￾tion, the PDT to UD conversion is relatively straight￾forward. However, coordination may form very in￾tricate structures when combined with ellipsis. 2.2.2. Ellipsis Due to the adopted restrictions (esp. not allowing extra nodes to be added to the graph), the rep￾resentation of ellipis is problematic in both frame￾works. W… view at source ↗
read the original abstract

Czech has been part of Universal Dependencies since its first release in 2015. It has also been one of the best represented languages, with the Prague Dependency Treebank being order of magnitude larger than most other UD treebanks. More recently, three other datasets from the Prague family were added and the annotations thoroughly revisited, forming the "Prague Dependency Treebank-Consolidated" (PDT-C). In comparison to the original PDT, PDT-C is more than twice as large, but it is also much more diverse in terms of genres and domains. In this paper, we describe the conversion of the new resource to Universal Dependencies. While the two annotation schemes are relatively similar at the first sight, there are numerous small differences in topology of the dependency structures and in granularity of the POS and relation type inventories. We demonstrate a selection of such differences on examples, discuss the diverging motivations, as well as ways to overcome the differences during conversion. We argue that while PDT is less "universal" and more tightly bound to one language, its multi-layer annotation is rich and provides all information needed for basic UD trees, and much more.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

1 major / 1 minor

Summary. The paper describes the conversion of the Prague Dependency Treebank-Consolidated (PDT-C) to Universal Dependencies, producing the UD_Czech-PDTC treebank. PDT-C is presented as more than twice the size of the original PDT and substantially more diverse in genres and domains. The authors identify differences in dependency topology, POS granularity, and relation inventories between PDT and UD, illustrate them with examples, discuss the underlying motivations, and outline conversion strategies. They argue that PDT's multi-layer annotation supplies all information required for basic UD trees (and more).

Significance. If the conversion is shown to be reliable, the resulting treebank would be a valuable addition to UD for Czech, offering substantially greater scale and genre coverage than prior Czech UD resources. This could improve model training and cross-domain evaluation for Czech NLP. The work is a standard resource contribution whose primary strength lies in the released data rather than novel theoretical claims.

major comments (1)
  1. [Abstract] Abstract: the central claim that differences in topology, POS, and relations 'can be systematically overcome during conversion without introducing inconsistencies or loss of essential information' is not supported by any quantitative validation, error analysis, sample-based accuracy figures, or inter-annotator agreement on the converted output. This directly bears on the weakest assumption identified in the review and must be addressed for the resource to be usable with confidence.
minor comments (1)
  1. The manuscript would benefit from an explicit table (or section) reporting the final token/sentence counts, genre breakdown, and comparison to existing Czech UD treebanks to quantify the claimed increase in size and diversity.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for the constructive feedback. We address the single major comment below and agree that additional validation is needed to support the claims about the conversion process.

read point-by-point responses

  1. Referee: [Abstract] Abstract: the central claim that differences in topology, POS, and relations 'can be systematically overcome during conversion without introducing inconsistencies or loss of essential information' is not supported by any quantitative validation, error analysis, sample-based accuracy figures, or inter-annotator agreement on the converted output. This directly bears on the weakest assumption identified in the review and must be addressed for the resource to be usable with confidence.

    Authors: We agree that the abstract claim would be strengthened by quantitative evidence. The conversion relies on deterministic, rule-based mappings from PDT-C's multi-layer annotations (morphology, syntax, tectogrammatics), which we argue supply the necessary information without loss. However, the current manuscript provides only qualitative examples and does not include error rates or sample validation. In the revised version, we will add a dedicated section with a manual check on a random sample of sentences (e.g., 500 sentences), reporting conversion accuracy, common error types, and any inconsistencies encountered. We will also revise the abstract to reference this validation. revision: yes

Circularity Check

0 steps flagged

No significant circularity detected

full rationale

The paper is a descriptive resource paper on converting the PDT-C treebank to UD format. It contains no equations, derivations, predictions, or self-referential claims that reduce inputs to outputs by construction. The central claim—that PDT-C's multi-layer annotation supplies all information needed for UD trees and that differences in topology/POS/relations can be bridged—is addressed directly by demonstrating examples and discussing conversion strategies, with no load-bearing self-citation chains or ansatzes imported from prior author work. The work is self-contained as a data release description.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

This is a descriptive resource paper; it introduces no free parameters, mathematical axioms, or invented entities.

pith-pipeline@v0.9.1-grok · 5768 in / 956 out tokens · 22379 ms · 2026-06-26T00:02:36.994764+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

31 extracted references · 7 canonical work pages

  1. [1]

    Haji c , Jan and Bej c ek, Eduard and B \'e mov \'a , Alevtina and Bur \'a n ov \'a , Eva and Fu c \' kov \'a , Eva and Haji c ov \'a , Eva and Havelka, Ji r \' and Hlav \'a c ov \'a , Jaroslava and Homola, Petr and Ircing, Pavel and K \'a rn \' k, Ji r \' and Kettnerov \'a , V \'a clava and Klyueva, Natalia and Kol \'a r ov \'a , Veronika and Ku c ov \'a...

    2024

  2. [2]

    Haji c , Jan and Panevov \' a , Jarmila and Haji c ov \' a , Eva and Sgall, Petr and Pajas, Petr and S t e p \' a nek, Jan and Havelka, Ji r \' i and Mikulov \' a , Marie and Z abokrtsk \' y , Zden e k and S ev c \' i kov \' a -Raz \' i mov \' a , Magda and Ure s ov \' a , Zde n ka. 2006. https://catalog.ldc.upenn.edu/LDC2006T01 Prague Dependency Treebank...

    2006

  3. [3]

    Nivre, Joakim and all. 2017. http://hdl.handle.net/11234/1-1983 Universal D ependencies 2.0 . LINDAT / CLARIAH - CZ digital library at the Institute of Formal and Applied Linguistics ( \'U FAL ), Charles University, Prague, Czech republic. PID http://hdl.handle.net/11234/1-1983 http://hdl.handle.net/11234/1-1983

    2017

  4. [4]

    Nivre, Joakim and Bosco, Cristina and Choi, Jinho and de Marneffe, Marie-Catherine and Dozat, Timothy and Farkas, Rich \'a rd and Foster, Jennifer and Ginter, Filip and Goldberg, Yoav and Haji c , Jan and Kanerva, Jenna and Laippala, Veronika and Lenci, Alessandro and Lynn, Teresa and Manning, Christopher and McDonald , Ryan and Missil \"a , Anna and Mont...

    2015

  5. [5]

    o bel, Nina and Bobicev, Victoria and Boizou, Lo \

    Zeman, Daniel and Nivre, Joakim and Abrams, Mitchell and Ackermann, Elia and Adolphe, Jephtey and Aepli, No \"e mi and Aghaei, Hamid and Agi \'c , Z eljko and Ahmadi, Amir and Ahrenberg, Lars and Ajede, Chika Kennedy and Akhundjanova, Arofat and Akkurt, Furkan and Aleksandravi c i \=u t \.e , Gabriel \.e and Alfina, Ika and Algom, Avner and Alnajjar, Khal...

    2025

  6. [6]

    o lker, Maximilian Wendt, Felix Hennig, and Arne K \

    Emanuel Borges V \"o lker, Maximilian Wendt, Felix Hennig, and Arne K \"o hn. 2019. https://doi.org/10.18653/v1/W19-8006 HDT - UD : A very large U niversal D ependencies treebank for G erman . In Proceedings of the Third Workshop on Universal Dependencies, pages 46--57, Paris, France. Association for Computational Linguistics

    work page doi:10.18653/v1/w19-8006 2019

  7. [7]

    Flavio Massimiliano Cecchini. 2024. https://doi.org/10.14712/00326585.029 Let's Do It Orderly: A Proposal for a Better Taxonomy of Adverbs in Universal Dependencies, and Beyond . The Prague Bulletin of Mathematical Linguistics, 121:15--65

    work page doi:10.14712/00326585.029 2024

  8. [8]

    Noam Chomsky. 1957. Syntactic Structures. Mouton, The Hague

    1957

  9. [9]

    Manning, Joakim Nivre, and Daniel Zeman

    Marie-Catherine de Marneffe, Christopher D. Manning, Joakim Nivre, and Daniel Zeman. 2021. https://doi.org/10.1162/coli_a_00402 U niversal D ependencies . Computational Linguistics, 47(2):255--308

    work page doi:10.1162/coli_a_00402 2021

  10. [10]

    Marie-Catherine de Marneffe, Joakim Nivre, and Daniel Zeman. 2024. https://www.linguisticanalysis.com/wp-content/uploads/2024/12/07-Function-Words-in-Universal-Dependencies-pp-549-628.pdf Function W ords in U niversal D ependencies . Linguistic Analysis, 43(3-4):549--588

    2024

  11. [11]

    Eva Haji c ov \'a , Marie Mikulov \'a , and Jarmila Panevov \'a . 2015. https://aclanthology.org/W15-2116/ Reconstructions of Deletions in a Dependency-based Description of C zech: Selected Issues . In Proceedings of the Third International Conference on Dependency Linguistics, pages 131--140. Uppsala University, Uppsala, Sweden

    2015

  12. [12]

    Eva Haji c ov \' a , Jarmila Panevov \' a , Marie Mikulov \' a , and Jan Haji c . 2024. https://www.linguisticanalysis.com/wp-content/uploads/2024/12/05-Function-Words-in-Praguian-Functional-Generative-Description-pp-465-512-.pdf Function W ords in P raguian F unctional G enerative D escription . Linguistic Analysis, 43(3-4):465--512

    2024

  13. [13]

    Max Jakob, Mark \'e ta Lopatkov \'a , and Valia Kordoni. 2010. https://aclanthology.org/L10-1342/ Mapping between Dependency Structures and Compositional Semantic Representations . In Proceedings of the Seventh International Conference on Language Resources and Evaluation, Valletta, Malta. European Language Resources Association

    2010

  14. [14]

    Mark \' e ta Lopatkov \' a , Eva Fu c \' kov \' a , Federica Gamba, Jan S t e p \' a nek, Daniel Zeman, and S \' a rka Zik \' a nov \' a . 2024. https://ceur-ws.org/Vol-3792/paper7.pdf Towards a Conversion of the Prague Dependency Treebank Data to the Uniform Meaning Representation . In Proceedings of the 24th Conference Information Technologies – Applica...

    2024

  15. [15]

    Marie Mikulov \'a . 2014. https://aclanthology.org/O14-1013/ Semantic Representation of Ellipsis in the P rague Dependency Treebanks . In Proceedings of the 26th Conference on Computational Linguistics and Speech Processing, pages 125--138, Jhongli, Taiwan. Association for Computational Linguistics and Chinese Language Processing

    2014

  16. [16]

    Marie Mikulov \'a , Ji r \' M \' rovsk \' y , Milan Straka, Pavl \'i na Synkov \'a , Barbora S t e p \'a nkov \'a , Jan S t e p \'a nek, and Jan Haji c . 2026. Prague Dependency Treebank - Consolidated 2.0: Enriching a Complex Annotation Scheme . In Proceedings of the Fifteenth Language Resources and Evaluation Conference, Palma de Mallorca, Spain. Europe...

    2026

  17. [17]

    Marie Mikulov \'a , Barbora S t e p \'a nkov \'a , and Jan S t e p \'a nek. 2025. https://aclanthology.org/2025.coling-main.147/ From F orm to M eaning: The C ase of P articles within the P rague D ependency T reebank A nnotation S cheme . In Proceedings of the 31st International Conference on Computational Linguistics, pages 2163--2175, Abu Dhabi, UAE. A...

    2025

  18. [18]

    Ji r \' M \' rovsk \' y and Pavl \'i na Synkov \'a . 2026. Presenting the P rague D iscourse T reebank 4.0. In Proceedings of the Fifteenth Language Resources and Evaluation Conference, Palma de Mallorca, Spain. European Language Resources Association

    2026

  19. [19]

    Anna Nedoluzhko, Michal Nov \'a k, Martin Popel, Zden e k Z abokrtsk \'y , Amir Zeldes, and Daniel Zeman. 2022. https://aclanthology.org/2022.lrec-1.520/ C oref UD 1.0: Coreference M eets U niversal D ependencies . In Proceedings of the Thirteenth Language Resources and Evaluation Conference, pages 4859--4872, Marseille, France. European Language Resource...

    2022

  20. [20]

    Manning, Sampo Pyysalo, Sebastian Schuster, Francis Tyers, and Daniel Zeman

    Joakim Nivre, Marie-Catherine de Marneffe, Filip Ginter, Jan Haji c , Christopher D. Manning, Sampo Pyysalo, Sebastian Schuster, Francis Tyers, and Daniel Zeman. 2020. https://aclanthology.org/2020.lrec-1.497/ U niversal D ependencies v2: An Evergrowing Multilingual Treebank Collection . In Proceedings of the Twelfth Language Resources and Evaluation Conf...

    2020

  21. [21]

    Stephan Oepen, Omri Abend, Lasha Abzianidze, Johan Bos, Jan Hajič, Daniel Hershcovich, Bin Li, Tim O ' Gorman, Nianwen Xue, and Daniel Zeman. 2020. https://doi.org/10.18653/v1/2020.conll-shared.1 MRP 2020: The Second Shared Task on Cross-Framework and Cross-Lingual Meaning Representation Parsing . In Proceedings of the CoNLL 2020 Shared Task: Cross-Framew...

    work page doi:10.18653/v1/2020.conll-shared.1 2020

  22. [22]

    Stephan Oepen, Omri Abend, Jan Hajič, Daniel Hershcovich, Marco Kuhlmann, Tim O ' Gorman, Nianwen Xue, Jayeol Chun, Milan Straka, and Zdeňka Urešová. 2019. https://doi.org/10.18653/v1/K19-2001 MRP 2019: Cross-Framework Meaning Representation Parsing . In Proceedings of the Shared Task on Cross-Framework Meaning Representation Parsing at the 2019 Conferenc...

    work page doi:10.18653/v1/k19-2001 2019

  23. [23]

    Martin Popel, David Mare c ek, Jan S t e p \'a nek, Daniel Zeman, and Zden e k Z abokrtsk \'y . 2013. https://aclanthology.org/P13-1051/ Coordination Structures in Dependency Treebanks . In Proceedings of the 51st Annual Meeting of the Association for Computational Linguistics, pages 517--527, Sofia, Bulgaria. Association for Computational Linguistics

    2013

  24. [24]

    Alexandr Rosen. 2023. https://www.juls.savba.sk/ediela/jc/2023/1/jc23-01.pdf#page=256 The InterCorp parallel corpus with a uniform annotation for all languages . Jazykovedn \' y c asopis / Journal of Linguistics , 74(1):254--265

    2023

  25. [25]

    Agata Rozumko. 2016. Linguistic Concepts across Languages: The Category of Epistemic Adverbs in English and Polish . Yearbook of the Poznan Linguistic Meeting, 2(1):195--214

    2016

  26. [26]

    Sebastian Schuster and Christopher D. Manning. 2016. https://aclanthology.org/L16-1376 Enhanced E nglish U niversal D ependencies: An Improved Representation for Natural Language Understanding Tasks . In Proceedings of the Tenth International Conference on Language Resources and Evaluation, pages 2371--2378, Portoro z , Slovenia. European Language Resourc...

    2016

  27. [27]

    Petr Sgall, Eva Haji c ov\' a , and Jarmila Panevov\' a . 1986. The Meaning of the Sentence and Its Semantic and Pragmatic Aspects . Academia/Reidel Publishing Company, Prague/Dordrecht

    1986

  28. [28]

    Jan S t e p \'a nek, Daniel Zeman, Mark \'e ta Lopatkov \'a , Federica Gamba, and Hana Hled \'i kov \'a . 2025. https://aclanthology.org/2025.dmr-1.1/ Comparing Manual and Automatic UMR s for C zech and L atin . In Proceedings of the Sixth International Workshop on Designing Meaning Representations, pages 1--12, Prague, Czechia. Association for Computatio...

    2025

  29. [29]

    Milan Straka. 2018. https://doi.org/10.18653/v1/K18-2020 UDP ipe 2.0 Prototype at C o NLL 2018 UD Shared Task . In Proceedings of the C o NLL 2018 Shared Task: Multilingual Parsing from Raw Text to Universal Dependencies , pages 197--207, Brussels, Belgium. Association for Computational Linguistics

    work page doi:10.18653/v1/k18-2020 2018

  30. [30]

    Milan Straka, Jana Strakov \'a , and Jan Haji c . 2019. https://doi.org/10.18653/v1/W19-4212 UDP ipe at SIGMORPHON 2019: Contextualized Embeddings, Regularization with Morphological Categories, Corpora Merging . In Proceedings of the 16th Workshop on Computational Research in Phonetics, Phonology, and Morphology, pages 95--103, Florence, Italy. Associatio...

    work page doi:10.18653/v1/w19-4212 2019

  31. [31]

    Daniel Zeman, Pavel Kosek, Martin Březina, and Jiří Pergler. 2023. https://www.juls.savba.sk/ediela/jc/2023/1/jc23-01.pdf#page=216 Morphosyntactic annotation in Universal Dependencies for old Czech . Jazykovedný časopis / Journal of Linguistics, 74(1):214--222

    2023