The Russian Legislative Corpus

Denis Saveliev; Ruslan Kuchakov

REVIEW 1 major objections 1 minor 5 references

The Russian Legislative Corpus

T0 review · 1 major / 1 minor · reviewed 2026-05-23 · grok-4.3

Challenge this review Download PDF Read on arXiv ↗

Pith's one-line read A corpus compiles 304,382 Russian legislative texts from 1991 to 2025, available in plain and linguistically annotated forms.

desk verdict This is a data release of a Russian legislative corpus with UD annotations, but the collection method is undocumented enough that the size claims can't be verified. read the letter →

arxiv 2406.04855 v3 submitted 2024-06-07 cs.CL

Denis Saveliev , Ruslan Kuchakov This is my paper · ORCID

classification cs.CL

keywords RussianlegislationlegalcorpusUniversalDependenciesCoNLL-Uformatnaturallanguageprocessing

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

The paper introduces a large collection of primary and secondary Russian laws covering the post-Soviet period through 2025. It supplies the full texts along with metadata in one version and adds part-of-speech, morphological, and syntactic annotations in Universal Dependencies format in the second version. The resource totals over 194 million tokens and is positioned for use in computational linguistics and legal studies. A sympathetic reader would see this as filling a gap in publicly available, machine-readable Russian legal data for downstream tasks such as parsing or historical analysis.

What carries the argument

The Russian Legislative Corpus itself, which supplies both raw texts and their conversion to Universal Dependencies CoNLL-U format for annotation.

What would settle it

Locating an official Russian law or regulation adopted between 1991 and 2025 that is absent from the corpus, or finding systematic duplicates or transcription mismatches when compared against government publication records.

Watch

Extended reading notes

Core claim

The authors have assembled and released a corpus of 304,382 Russian legislative documents (194,425,905 tokens) spanning 1991 to 2025, offered in a basic form with metadata and a detailed form that includes the original texts plus their CoNLL-U equivalents annotated for parts of speech, morphology, and syntactic dependencies.

Load-bearing premise

The collected documents accurately and exhaustively cover all primary and secondary Russian legislation in the period without major omissions, duplicates, or transcription errors.

Editorial extensions

If this is right

Researchers can now run large-scale statistical or machine-learning studies on the language and structure of Russian statutes over three decades.
The annotated version supports training or evaluation of dependency parsers and morphological analyzers on legal-domain Russian text.
Temporal subsets of the corpus enable tracking of changes in legislative phrasing or complexity across years.
The resource can serve as a test bed for information-extraction systems aimed at legal documents.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

Cross-referencing the corpus against official Russian government portals could quantify coverage gaps that the paper does not report.
The same texts could be aligned with parallel corpora from other languages to study translation patterns in international law.
Periodic updates to the corpus would allow longitudinal modeling of how new legislation interacts with prior texts.

Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit.

Desk Editor's Note

read the letter

The main thing to know is that the authors have put together a corpus of Russian primary and secondary legislation from 1991 to 2025, with 304k texts and 194 million tokens, and one version comes with full Universal Dependencies parsing in CoNLL-U format including POS, morphology, and dependencies. That parsed version is the practical output here. It looks new at this scale for Russian legal text under UD, since earlier resources the paper cites are smaller or lack the syntactic layer. If the numbers are accurate, it gives legal NLP researchers working in Russian a ready-made annotated set they can use without starting from scratch on parsing. The collection step is the clear weak point. The headline claim of comprehensiveness depends on having pulled every relevant text without big gaps or duplicates, yet the description gives no scraping protocol, no year-by-year comparison to official registries, and no checks for errors or overlap. That leaves the central numbers untestable from the outside. If the full paper adds a reproducible method section and some validation stats, that would fix most of the issue; otherwise users will need to run their own audits. This is for people in Russian computational linguistics or legal NLP who need domain data rather than new methods. A reader who wants annotated legal text to train or evaluate models would get direct value. It deserves peer review as a resource paper, since the underlying idea is sound and the format is usable, even if the documentation needs tightening. I would send it out with a request for the missing collection and validation details.

Referee Report

1 major / 1 minor

Summary. The manuscript presents the Russian Legislative Corpus, a collection of 304,382 primary and secondary legislative texts adopted between 1991 and 2025, totaling 194,425,905 tokens. Two versions are released: a basic version with simple metadata and a detailed version containing the original texts plus Universal Dependencies CoNLL-U annotations for POS tags, morphological features, and syntactic dependencies.

Significance. A verified exhaustive corpus of this scale would constitute a substantial resource for Russian NLP, enabling large-scale studies of legislative language, diachronic analysis, and training of domain-specific models. The inclusion of UD annotations further increases its utility for syntactic and morphological research. The significance, however, depends entirely on the undocumented collection and validation steps.

major comments (1)

[Abstract / Data Collection (absent)] Abstract and entire manuscript: the headline claim that the 304,382 texts constitute a comprehensive, non-duplicative record of all Russian primary and secondary legislation 1991–2025 is unsupported by any description of data sources, scraping or API protocols, year-by-year benchmarking against official registries, deduplication methods, or sampling-based error audits. Without these, the numerical claims cannot be independently verified and the central contribution remains untestable.

minor comments (1)

[Abstract] The abstract states token counts but provides no breakdown by year, document type, or source; adding such tables would improve transparency even if full methodology is added elsewhere.

Simulated Author's Rebuttal

1 responses · 0 unresolved

Thank you for the thorough review. We agree that the data collection methodology requires more detailed documentation to support the claims of comprehensiveness. We will update the manuscript with a new section on corpus construction.

read point-by-point responses

Referee: Abstract and entire manuscript: the headline claim that the 304,382 texts constitute a comprehensive, non-duplicative record of all Russian primary and secondary legislation 1991–2025 is unsupported by any description of data sources, scraping or API protocols, year-by-year benchmarking against official registries, deduplication methods, or sampling-based error audits. Without these, the numerical claims cannot be independently verified and the central contribution remains untestable.

Authors: We will revise the manuscript to include a detailed 'Data Collection' section that describes the primary sources (official Russian Federation legal information portals), the automated collection protocols, deduplication procedures based on unique identifiers and content hashing, and cross-validation with annual official reports on legislative activity. This will substantiate the corpus's scope and allow for independent verification. revision: yes

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: data release paper with no derivations or fitted results

full rationale

The paper is a corpus presentation whose central claim is the existence, size (304,382 texts), and availability of the Russian legislative corpus. No equations, predictions, first-principles derivations, or fitted parameters are present. The collection is described as sourced from official portals but involves no mathematical reduction or self-citation chain that could be circular. The exhaustiveness claim is a factual assertion about data ingestion, not a derived result equivalent to its inputs by construction. This is the normal non-finding for a data-release manuscript.

Assumptions & free parameters 0 free parameters · 0 assumptions · 0 invented entities

No mathematical model, free parameters, or invented entities are introduced; the paper is a straightforward data collection and annotation effort relying on standard NLP tooling.

how reviews work

0 comments

Cite this review

Pith. "Pith review of The Russian Legislative Corpus." pith.science (2026). https://pith.science/paper/2406.04855

@misc{pith2026240604855,
  author       = {Pith},
  title        = {Pith review of: The Russian Legislative Corpus},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/2406.04855}},
  note         = {Machine review of arXiv:2406.04855}
}

read the original abstract

We present a comprehensive corpus of Russian primary and secondary legislation adopted between 1991 and 2025, comprising 304,382 texts (194,425,905 tokens). The corpus is available in two versions: the basic version contains texts with simple metadata, while the detailed version includes both the original texts and their equivalents converted to the Universal Dependencies CoNLL-U format, annotated with parts of speech, morphological features, and syntactic dependencies.

Figures

Figures reproduced from arXiv: 2406.04855 by the authors.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

5 extracted references · 5 canonical work pages

[1]

Corpus Linguistics–2021

Blinova, O. V. and N. A. Tarasov (2021). Complexity of Russian legal texts: assessment methods and language data. In Proceedings of the International Conference "Corpus Linguistics–2021" . https://events.spbu.ru/eventsContent/events/2021/corpora/Корпусная

work page 2021
[2]

Blinova, O. V. and N. A. Tarasov (2023). Language complexity across sub-styles and genres in legal Russian . Research Result. Theoretical and Applied Linguistics\/ 9\/ (2), 73--96. 10.18413/2313-8912-2023-9-2-0-5

work page doi:10.18413/2313-8912-2023-9-2-0-5 2023
[3]

Kuchakov, R. and D. Saveliev (2018). The complexity of the Russian legislation from the lexical and syntactic perspective ( Slozhnost' pravovyh aktov v Rossii : leksicheskoe i sintaksicheskoe kachestvo tekstov). Analytic report, IRL European University\/ . https://enforce.spb.ru/images/analit_zapiski/memo_readability_2018_web.pdf (in Russian)

work page 2018
[4]

Saveliev, D. (2018). On creating and using text of the Russian Federation corpus of legal acts acts as open dataset. Pravo. Zhurnal Vysshey shkoly ekonomiki\/ (1), 26--44. 10.17323/2072-8166.2018.1.26.44 (in Russian)

work page doi:10.17323/2072-8166.2018.1.26.44 2018
[5]

Saveliev, D. (2020). A study in complexity of sentences constituting Russian Federation legal acts. Pravo. Zhurnal Vysshey shkoly ekonomiki\/ (1), 50--74. 10.17323/2072-8166.2020.1.50.74 (in Russian)

work page doi:10.17323/2072-8166.2020.1.50.74 2020

Pith tools

Reviewed May 23, 2026 · model on record in the stance chip above.

[1] [1]

Corpus Linguistics–2021

Blinova, O. V. and N. A. Tarasov (2021). Complexity of Russian legal texts: assessment methods and language data. In Proceedings of the International Conference "Corpus Linguistics–2021" . https://events.spbu.ru/eventsContent/events/2021/corpora/Корпусная

work page 2021

[2] [2]

Blinova, O. V. and N. A. Tarasov (2023). Language complexity across sub-styles and genres in legal Russian . Research Result. Theoretical and Applied Linguistics\/ 9\/ (2), 73--96. 10.18413/2313-8912-2023-9-2-0-5

work page doi:10.18413/2313-8912-2023-9-2-0-5 2023

[3] [3]

Kuchakov, R. and D. Saveliev (2018). The complexity of the Russian legislation from the lexical and syntactic perspective ( Slozhnost' pravovyh aktov v Rossii : leksicheskoe i sintaksicheskoe kachestvo tekstov). Analytic report, IRL European University\/ . https://enforce.spb.ru/images/analit_zapiski/memo_readability_2018_web.pdf (in Russian)

work page 2018

[4] [4]

Saveliev, D. (2018). On creating and using text of the Russian Federation corpus of legal acts acts as open dataset. Pravo. Zhurnal Vysshey shkoly ekonomiki\/ (1), 26--44. 10.17323/2072-8166.2018.1.26.44 (in Russian)

work page doi:10.17323/2072-8166.2018.1.26.44 2018

[5] [5]

Saveliev, D. (2020). A study in complexity of sentences constituting Russian Federation legal acts. Pravo. Zhurnal Vysshey shkoly ekonomiki\/ (1), 50--74. 10.17323/2072-8166.2020.1.50.74 (in Russian)

work page doi:10.17323/2072-8166.2020.1.50.74 2020