Pith. sign in

REVIEW 3 major objections 5 minor 44 references

Computational Identification of Regulatory Statements in EU Legislation

T0 review · 3 major / 5 minor · reviewed 2026-08-16 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read Two automated methods — one grammar-based, one transformer-based — identify regulatory statements in EU legislation with 80–84% accuracy, and their low agreement points toward a better hybrid.

desk verdict A workmanlike empirical comparison of two classifiers for regulatory sentences in EU law, with a genuinely new corpus and open code/data, but the missing inter-annotator agreement on the gold standard makes the headline accuracy numbers less secure than they look. read the letter →

arxiv 2505.00479 v1 pith:HCHBABCB submitted 2025-05-01 cs.CL

Gijs Jan Brandsma , Jens Blom-Hansen , Christiaan Meijer , Kody Moodley This is my paper · ORCID
classification cs.CL
keywords regulatorystatementsEUlegislationinstitutionalgrammartooltextclassificationdependencyparsingtransferlearninglegalexplainableAI
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 attacks a bottleneck for measuring how dense and strict EU law has become: deciding which sentences actually impose a legal rule on a specific actor, as opposed to conditions, amendments, or boilerplate about entry into force. It fixes a definition of a regulatory statement using the Institutional Grammar Tool's ADICO scheme, requiring an attribute expressed as an agent noun, a deontic auxiliary such as 'must' or 'shall', and an aim. Two classifiers are built to that definition, one from hand-written dependency-parsing rules and one from a transformer pretrained on legal text with a tree-boosted classifier on top. On a new 7,200-sentence corpus stratified across policy areas and years, both reach accuracies above 0.80 (0.80 and 0.84). Since the two models agree only moderately (Krippendorff's alpha 0.58), the paper argues that a hybrid using the grammar rules to certify attributes and the transformer to catch implicit or unfamiliar actors should do better.

What carries the argument

The load-bearing mechanism is the ADICO decomposition from the Institutional Grammar Tool, cut down to ADI: every regulatory sentence must name the actor (Attribute) as an agent noun, carry a deontic auxiliary ('must' or 'shall'), and state an action (Aim). The dependency approach operationalizes ADI as a small set of parse-tree rules — a lexical verb whose auxiliary is deontic, with a path through the tree to an agent noun checked against a knowledge graph. The transformer approach operationalizes the same definition statistically by training a shallow classifier on sentence features from a legal-text pretrained transformer. The two mechanisms are deliberately orthogonal, and their disagreement is treated as reusable signal rather than noise.

What would settle it

Have both annotators independently label the same random sample of a few hundred sentences from the corpus and compute Krippendorff's alpha between them, and also implement the proposed grammar-with-transformer-fallback hybrid on the released test set; low annotator agreement would undermine the accuracy numbers, and a hybrid that fails to beat 0.84 would undermine the complementary-strengths claim.

Watch

Extended reading notes

Core claim

The paper's central claim is that the notion of a regulatory statement can be pinned down by an IGT-based definition — a sentence must explicitly carry the Attribute, Deontic, and Aim components, with the attribute realized as an agent noun — and that this definition can be learned by two very different machines. The dependency approach encodes the definition directly as parse-tree conditions: a lexical verb governed by a 'must' or 'shall' auxiliary, plus a dependency path from that verb to an agent noun, where agenthood is decided with the help of a knowledge graph. The transfer-learning approach instead extracts sentence features from a legal-domain transformer and classifies them with gradient-boosted trees. On a stratified corpus of 7,200 sentences drawn from roughly 120,000 EU legal acts across 1971-2022, both methods surpass 0.80 accuracy, the transformer reaching 0.84. The grammar rules are more precise on the regulatory class while the transformer has higher regulatory recall, and the model-model agreement is only 0.58; those two facts are the paper's evidence that the methods capture different linguistic signals.

Load-bearing premise

The ground-truth labels for the 7,200 sentences were each produced by one of two annotators who split the batch, and no inter-annotator agreement is reported; if the two scholars interpret the institutional-grammar definition differently, the reported 0.80 and 0.84 accuracy figures are not reliable as absolute numbers.

Editorial extensions

If this is right

  • Regulatory density and strictness can now be measured automatically over the full corpus of approximately 180,000 EU legal acts, broken down by year and policy area.
  • A sentence flagged as regulatory by the dependency rules is highly trustworthy (high precision), while the transformer recovers regulatory sentences whose actor is implicit or whose name the grammar rules do not know.
  • Because the error sets are largely disjoint, a concrete hybrid — let the grammar rules identify the attribute, and fall back to the transformer when they fail — is a direct next step that the paper argues should improve accuracy.
  • The explanation analysis indicates that the transformer keys on action verbs for regulatory sentences and on amendment and context lexicon (such as 'replaced by the following') for non-regulatory ones, giving an interpretable handle on its decisions.
  • The code, data, and models are released publicly, so the comparison can be reproduced, audited, and extended by other groups.

Reading between the lines

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

  • If the hybrid direction is right, a confidence-weighted ensemble over the two classifiers is a natural variant; since both models and the test set are released, this can be checked directly without new data collection.
  • The single-annotator ground truth is the main fragility; an immediate validation experiment would have both scholars independently label a shared random subsample, since the paper never reports annotator-annotator agreement.
  • The choice to count only core ADI sentences means that amendments, preconditions, and or-else sanctions are excluded from the density metric; in heavily amended policy areas, regulatory strictness may therefore be understated, and a complementary metric counting those clauses would test how much this matters.
  • For the pre-1990 documents, OCR-induced sentence splicing is acknowledged as a corruption source; applying the trained models to the earliest decades of the corpus will require a layout-aware sentence segmenter, and the authors' expectation that results generalize to those documents is a testable assumption.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 5 minor

Summary. The paper addresses the task of automatically identifying regulatory statements in EU legislation, framed as binary sentence classification. It defines a regulatory statement using the Institutional Grammar Tool (IGT), requiring at least Attribute-Deontic-Aim components with the attribute expressed as an agent noun. The authors construct a corpus of about 7,200 human-labeled sentences sampled from roughly 120,000 EU legislative documents, stratified by adoption year and policy area. They compare two approaches: a hand-crafted dependency-parsing rule set built on SpaCy and ConceptNet, and a transfer-learning classifier using LEGAL-BERT features fed to XGBoost. They report accuracies of 0.80 and 0.84, respectively, a between-model Krippendorff alpha of 0.58, and provide a LIME-based XAI analysis. The paper argues that the two methods have complementary strengths and proposes a fusion strategy, while also discussing limitations such as OCR artifacts and attribute identification difficulties.

Significance. If the reported results are robust, the paper would make a useful contribution to computational legal-text analysis: it provides a concrete, IGT-based definition of regulatory statements, a sizable publicly available labeled corpus, and a comparison of a grammar-based method with a transformer-based method on the same test set. The reproducibility effort is a clear strength: code repositories, raw data, and processed data are made available. The hand-crafted dependency rules are tested on a separate holdout set rather than tuned on it, which is methodologically sound. The XAI analysis, though preliminary, is a useful step toward interpreting the transformer model. However, the central empirical claims are currently not fully supported because the ground truth was produced without measured inter-annotator reliability and the test set covers only a limited time period relative to the claimed scope.

major comments (3)
  1. [Section 3.4 and Section 4.1] The main empirical claim—that both classifiers reach 0.80/0.84 accuracy and that their 0.58 agreement supports a hybrid—rests on a gold standard whose reliability is not established. Section 3.4 states that the two annotators split the 7,200-sentence batch into two equal halves and labeled them separately, with no overlapping annotation and no reported inter-annotator agreement. Because the train/test split is random over the pooled halves, the test labels are a mixture of two potentially different interpretations of the IGT definition, and the reported accuracy values and the between-model Krippendorff alpha of 0.58 conflate model error with annotator disagreement. The authors should either add an overlapping annotation sample and report agreement, or explicitly reframe all headline numbers as conditional on a single-annotator label convention. This issue is load-bearing for the abstract and for Section 4.1.
  2. [Section 5 (OCR issue) and Section 3.2] The claim that the classifiers are applicable to the full 1970–2022 corpus is not supported by the evaluation data. Section 5 states that, because of OCR splicing errors, sentences for evaluation were sampled primarily from 2000–2023; Section 3.2, however, states that 2023 and 2024 were excluded and that the corpus covers 1970–2022. Taken together, the test set contains essentially no pre-2000 sentences, so the reported test-set accuracy cannot validate generalization to pre-1990 documents. The statement in Section 5 that manual inspection showed no significant linguistic differences is an anecdotal justification and should be either backed by a quantitative comparison or replaced by a narrower claim about post-2000 legislation.
  3. [Section 4.1, Table 1] Table 1 reports point estimates with no confidence intervals or significance tests. With roughly 1,450 test sentences, the difference between 0.80 and 0.84 may be within sampling variability, and the claim that both approaches 'performed similarly well' is not backed by any statistical comparison. Similarly, the 0.58 between-model alpha is interpreted as evidence of complementary strengths, but a low agreement metric alone is not a direct measure of error complementarity; a direct analysis of the sets of misclassified sentences, or an oracle-combination experiment, would be needed to substantiate the complementarity claim.
minor comments (5)
  1. [Section 2] The citation for BERT appears to be wrong: 'Bidirectional Encoder Representations from Transformers (BERT) [8]' should cite Devlin et al. [7], not Dragoni et al. [8].
  2. [Section 3.4] The text says 'approximately 7,200 sentences,' but the described equal allocation implies 1,034 strata × 7 sentences = 7,238 sentences; please state the exact number and clarify the discrepancy.
  3. [Abstract and Section 3.2] The abstract cites approximately 180,000 published legal acts between 1952 and 2023, while Section 3.2 reports roughly 120,000 documents between 1970 and 2022; please reconcile these numbers or explain the different scopes.
  4. [Section 4.3] The proposed fusion strategy is described as benefiting from both methods' strengths, and Section 7 states it 'could significantly improve performance,' but the strategy is not evaluated; please either add a small experiment or soften the wording.
  5. [Table 1 caption] The table reports Krippendorff alpha values for agreement with ground truth, while the text separately reports alpha = 0.58 for inter-model agreement; the table caption should clarify this distinction to avoid confusion.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: both classifiers are evaluated against independently held-out human labels, and no fitted value is renamed as a prediction.

full rationale

The paper's central empirical claims are the classification accuracies of two approaches (0.80 and 0.84) and their inter-model agreement (Krippendorff alpha = 0.58). These are measured against a held-out test set of 20% of the 7,200-sentence corpus, which is separate from the 80% training set used for the transformer/XGBoost model and is also used to evaluate the dependency-parsing rules. The dependency rules were initially developed on a separate 100-sentence sample and then applied to the larger held-out test set, so their performance is not fit-by-construction to the evaluation data. The transformer model's predictions are likewise a genuine out-of-sample evaluation, not a restatement of its training labels. The IGT-based definition of a regulatory statement is an operational definition used to create the gold standard; using that standard to measure accuracy is standard supervised evaluation, not circular reasoning. The inference from K alpha = 0.58 that the two models capture complementary patterns is an empirical observation about the two models' outputs, not a derivation from the models' inputs. The only self-citation is the DIANNA XAI library (Ranguelova et al. 2022, which includes co-author Meijer), but DIANNA is used as an off-the-shelf analysis tool for a preliminary interpretability study and does not supply or justify the paper's main classification results. Concerns about the gold standard resting on a single annotator per sentence and the absence of inter-annotator agreement are validity and reproducibility limitations, not circularity. Therefore, no step in the paper's derivation chain reduces to its own inputs by definition, by fitted-parameter renaming, or by load-bearing self-citation.

Assumptions & free parameters 0 free parameters · 5 assumptions · 0 invented entities

The paper makes no parametric circularity, but its central results rest on several domain assumptions about how EU law is written and how the IGT definition is operationalized. The most fragile assumption is that the single-annotator labels constitute reliable ground truth.

assumptions (5)
  • domain assumption The IGT ADICO framework, specifically that a regulatory statement must contain ADI components with an agent noun, correctly captures what should count as a regulatory statement for density and strictness metrics.
    This is the paper's own task definition; if it is not the right operationalization, the entire classification task and downstream metrics are misaligned. Section 3.1.
  • domain assumption Regulatory statements in EU legislation are adequately captured by sentences containing the deontic phrases 'shall' or 'must'.
    Section 3.2 filters the 0.6 million candidate sentences to those containing 'shall' or 'must', excluding other deontic expressions like 'may' by design.
  • domain assumption The two annotators' labels constitute a reliable ground truth for the IGT-based definition.
    Section 3.4 describes each sentence being labeled by one of two annotators, with no inter-annotator agreement statistic reported. If the annotators systematically disagree, the accuracy numbers are unstable.
  • domain assumption EU legislative documents have consistent structural markers that identify the boundaries of the regulatory (articles) section.
    Section 3.2 relies on a dictionary of start phrases like 'HAS ADOPTED THIS REGULATION' and end phrases like 'Done at Brussels' to extract candidate sentences; this consistency is assumed across the corpus.
  • domain assumption ConceptNet's hyponym relation to the 'agent' class adequately identifies agent nouns for the ADI definition.
    Section 3.3 uses a ConceptNet dictionary lookup to identify agent nouns, which is an imperfect proxy for the linguistic category of agent nouns, as the paper itself notes for abbreviations and unseen terms.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Computational Identification of Regulatory Statements in EU Legislation." pith.science (2026). https://pith.science/paper/HCHBABCB

@misc{pith2026250500479,
  author       = {Pith},
  title        = {Pith review of: Computational Identification of Regulatory Statements in EU Legislation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/HCHBABCB}},
  note         = {Machine review of arXiv:2505.00479}
}
read the original abstract

Identifying regulatory statements in legislation is useful for developing metrics to measure the regulatory density and strictness of legislation. A computational method is valuable for scaling the identification of such statements from a growing body of EU legislation, constituting approximately 180,000 published legal acts between 1952 and 2023. Past work on extraction of these statements varies in the permissiveness of their definitions for what constitutes a regulatory statement. In this work, we provide a specific definition for our purposes based on the institutional grammar tool. We develop and compare two contrasting approaches for automatically identifying such statements in EU legislation, one based on dependency parsing, and the other on a transformer-based machine learning model. We found both approaches performed similarly well with accuracies of 80% and 84% respectively and a K alpha of 0.58. The high accuracies and not exceedingly high agreement suggests potential for combining strengths of both approaches.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

44 extracted references · 20 canonical work pages

  1. [1]

    Computational identification of regulatory statements in EU legislation Gijs Jan Brandsmaa Jens Blom-Hansenb Christiaan Meijerc Kody Moodley1,c,d a Institute of Management Research, Radboud University, Nijmegen, The Netherlands b Aarhus University, Aarhus, Denmark c The Netherlands eScience Center, Amsterdam, The Netherlands d ORCID: 0000-0001-5666-1658 g...

    work page 1952

  2. [2]

    United Kingdom

    expresses the attribute (A) component as an agent noun where an agent noun is a noun that describes a person or thing that performs an action. The concept of agent noun subsumes proper nouns computationally identifiable through part of speech (POS) tagging e.g., “United Kingdom”. But it also includes the class of nouns for which the classification problem...

    work page 1970

  3. [3]

    It” and “ey

    Example regulatory sentences misclassified as non-regulatory by the dep-parser and correctly classified by the transformer model, with reasons for dep-parser misclassification. In Section 3.3, the second criteria for classifying a regulatory statement requires identification of an agent noun. However, in many cases this step is difficult due to several limit...

    work page 1990

  4. [4]

    It is customary to require α ≥ .800 (for statistically significant reliability). Where tentative conclusions are still acceptable, α ≥ .667 is the lowest conceivable limit

    Transfer learning model setup. Features are extracted from each sentence in our dataset, after which a shallow classifier was trained using the extracted features. We requested two human annotators (the researchers mentioned in Section 3.2) to generate new labelled examples using the following approach. To ensure that our sampled sentences were representa...

    work page 1971

  5. [7]

    Citizens must separate their recyclables

    which were not subject to this anomaly. Excluding these years from consideration obviously affected the generalizability of our results. However, human analysis of the language and phrasing used before 2000 in the regulations showed no significant differences and we therefore expect our results to generalize to pre-1990s documents. XAI methods: state-of-the-...

    work page 2000

  6. [8]

    Citizens must separate

    one cannot infer if there is a specific combination of these words which is a clearer indicator of the classification result. e theory is that if we could assign aribution scores to phrases e.g., “Citizens must separate” then we might be able to obtain a higher aribution score than 0.55 for this combined phrase than the individual words. And this would l...

    work page

  7. [15]

    arXiv:1810.04805

    Bert: Pre-training of deep bidirectional transformers for language understanding. arXiv:1810.04805. Retrieved from: https://arxiv.org/abs/1810.04805

    arXiv

  8. [16]

    In 1st Workshop on MIning and REasoning with Legal texts (MIREL 2016)

    Combining NLP approaches for rule extraction from legal documents. In 1st Workshop on MIning and REasoning with Legal texts (MIREL 2016). https://hdl.handle.net/11582/324998

    work page 2016

Show all 44 references
  1. [18]

    Artificial Intelligence in Law

    Natural language processing for legal document review: categorising deontic modalities in contracts. Artificial Intelligence in Law. https://doi.org/10.1007/s10506-023-09379-2

    doi

  2. [19]

    In Proceedings of the 2015 Conference on Empirical Methods in Natural Language Processing, 1373–1378

    An Improved Non-monotonic Transition System for Dependency Parsing. In Proceedings of the 2015 Conference on Empirical Methods in Natural Language Processing, 1373–1378. https://doi.org/10.18653/v1/D15-1162

    2015 doi

  3. [20]

    arXiv:1508.01991

    Bidirectional LSTM-CRF models for sequence tagging. arXiv:1508.01991. Retrieved from: https://arxiv.org/abs/1508.01991

    arXiv

  4. [21]

    Morphology 30, 185-218

    What’s in an agent? A distributional semantics approach to agent nouns in French. Morphology 30, 185-218. https://doi.org/10.1007/s11525-020-09366-2

    doi

  5. [24]

    https://doi.org/10.3233/FAIA220467

    362, 200-205. https://doi.org/10.3233/FAIA220467

    doi

  6. [25]

    following

    of the words in Figure 5 play the role of verbs indicating an action that forms part of the aim of the legal rule according to the ADICO framework. In contrast, there are no verb roles in Figure 6 for the non-regulatory classifications. If we hypothesize a pattern for Figure 6...

    2011

  7. [26]

    Artificial Intelligence and Law 27, 117-139

    CLAUDETTE: an automated detector of potentially unfair clauses in online terms of service. Artificial Intelligence and Law 27, 117-139. https://doi.org/10.1007/s10506-019-09243-2

    doi

  8. [27]

    arXiv:1907.11692

    Roberta: A robustly optimized bert pretraining approach. arXiv:1907.11692. Retrieved from: https://arxiv.org/abs/1907.11692

    1907 arXiv

  9. [29]

    International Conference on Learning Representations

    Efficient Estimation of Word Representations in Vector Space. International Conference on Learning Representations. https://doi.org/10.48550/arXiv.1301.3781

    doi arXiv

  10. [30]

    In Proceedings of the International Conference on Artificial Intelligence and Law (ICAIL 2017), 159-168

    Classifying sentential modality in legal language: a use case in financial regulations, acts and directives. In Proceedings of the International Conference on Artificial Intelligence and Law (ICAIL 2017), 159-168. https://doi.org/10.1145/3086512.308652

    2017

  11. [32]

    Legislative XML for the Semantic Web: Principles, Models, Standards for Document Management, 75-100

    Akoma-Ntoso for legal documents. Legislative XML for the Semantic Web: Principles, Models, Standards for Document Management, 75-100. https://doi.org/10.1007/978-94-007-1887-6_6

    doi

  12. [33]

    In Proceedings of 19th International Conference on Artificial Intelligence and Law (ICAIL 2023), 187-196

    Pre-trained Language Models for the Legal Domain: A Case Study on Indian Law. In Proceedings of 19th International Conference on Artificial Intelligence and Law (ICAIL 2023), 187-196. https://doi.org/10.1145/3594536.359516

    2023

  13. [35]

    In Proceedings of the British Machine Vision Conference (BMVC)

    RISE: Randomized Input Sampling for Explanation of Black-box Models. In Proceedings of the British Machine Vision Conference (BMVC). https://doi.org/10.48550/arXiv.1806.07421

    doi arXiv

  14. [37]

    In Proceedings of the 2016 Conference of the North American Chapter of the Association for Computational Linguistics: Demonstrations

    Why Should I Trust You? Explaining the Predictions of Any Classifier. In Proceedings of the 2016 Conference of the North American Chapter of the Association for Computational Linguistics: Demonstrations. San Diego, California: Association for Computational Linguistics, 97–101....

    2016 doi

  15. [38]

    arXiv:2211.12752

    Agent-specific deontic modality detection in legal language. arXiv:2211.12752. Retrieved from: https://arxiv.org/abs/2211.12752

    arXiv

  16. [39]

    arXiv:1910.01108

    DistilBERT, a distilled version of BERT: smaller, faster, cheaper and lighter. arXiv:1910.01108. Retrieved from: https://arxiv.org/abs/1910.01108

    1910 arXiv

  17. [40]

    https://doi.org/10.1609/aaai.v31i1.11164

    doi

  18. [41]

    Information Processing and Management 60(4):103374

    A BERT-based deontic logic learner. Information Processing and Management 60(4):103374. https://doi.org/10.1016/j.ipm.2023.103374

    2023

  19. [42]

    https://doi.org/10.3233/978-1-61499-609-5-189

    279, 189-190. https://doi.org/10.3233/978-1-61499-609-5-189

    doi

  20. [51]

    https://doi.org/10.1016/j.clsr.2023.105864

    2023

  21. [1995]

    American political science review 89, 3, 582-600

    A grammar of institutions. American political science review 89, 3, 582-600. https://doi.org/10.2307/2082975

    doi

  22. [2005]

    Princeton University Press

    Understanding Institutional Diversity. Princeton University Press. https://doi.org/10.2307/j.ctt7s7wm

    doi

  23. [2011]

    Springer Berlin Heidelberg

    Part-of-speech tagging from 97% to 100%: is it time for some linguistics? In International conference on intelligent text processing and computational linguistics, 171-189. Springer Berlin Heidelberg. https://doi.org/10.1007/978-3-642-19400-9_14

    doi

  24. [2013]

    In Proceedings of the International Semantic Web Conference (ISWC 2013), 151-166

    One license to compose them all: a deontic logic approach to data licensing on the web of data. In Proceedings of the International Semantic Web Conference (ISWC 2013), 151-166. Springer. https://doi.org/10.1007/978-3-642-41335-3_10

    2013 doi

  25. [2014]

    In Proceedings of the 2014 conference on empirical methods in natural language processing (EMNLP), 1532-1543

    Glove: Global vectors for word representation. In Proceedings of the 2014 conference on empirical methods in natural language processing (EMNLP), 1532-1543. https://doi.org/10.3115/v1/D14-1162

    2014 doi

  26. [2015]

    Reasoning Web

    LegalRuleML: Design principles and foundations. Reasoning Web. Web Logic Rules: 11th International Summer School, 151-188. https://doi.org/10.1007/978-3-319-21768-0_6

    doi

  27. [2016]

    https://doi.org/10.1145/2939672.2939785

    XGBoost: A Scalable Tree Boosting System, Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining. https://doi.org/10.1145/2939672.2939785

  28. [2017]

    In Proceedings of the 31st International Conference on Neural Information Processing Systems

    A unified approach to interpreting model predictions. In Proceedings of the 31st International Conference on Neural Information Processing Systems. Red Hook, NY, USA: Curran Associates Inc. (NIPS’17), 4768–4777. https://doi.org/10.48550/arXiv.1705.07874

    doi arXiv

  29. [2018]

    arXiv:1805.03871

    Obligation and prohibition extraction using hierarchical RNNs. arXiv:1805.03871. Retrieved from: hps://arxiv.org/abs/1805.03871

    arXiv

  30. [2019]

    In Proceedings of the Natural Legal Language Processing Workshop

    Extreme Multi-Label Legal Text Classification: A Case Study in EU Legislation. In Proceedings of the Natural Legal Language Processing Workshop. 78-87. https://doi.org/10.18653/v1/W19-2209

    doi

  31. [2020]

    arXiv:2010.02559

    LEGAL-BERT: The muppets straight out of law school. arXiv:2010.02559. Retrieved from: https://arxiv.org/abs/2010.02559

    2010 arXiv

  32. [2021]

    In Proceedings of the 29th ACM joint meeting on European software engineering conference and symposium on the foundations of software engineering, 1275-1280

    Domain adaptation for an automated classification of deontic modalities in software engineering contracts. In Proceedings of the 29th ACM joint meeting on European software engineering conference and symposium on the foundations of software engineering, 1275-1280. https://doi....

  33. [2022]

    In Proceedings of SAI Intelligent Systems Conference, 54-73

    Deontic Sentence Classification Using Tree Kernel Classifiers. In Proceedings of SAI Intelligent Systems Conference, 54-73. https://doi.org/10.1007/978-3-031-16072-1_4

    doi

  34. [2023]

    must” and “shall

    Past work on extraction of these statements varies in the permissiveness of their definitions for what constitutes a regulatory statement. In this work, we provide a specific definition for our purposes based on the institutional grammar tool (IGT). We then develop and compare tw...

    1952

  35. [2024]

    University of Luxembourg Law

    The Interplay Between Lawfulness and Explainability in the Automated Decisionmaking of the EU Administration. University of Luxembourg Law. Research Paper No. 2023-12. http://dx.doi.org/10.2139/ssrn.4561012

    2023 doi

  36. [4493]

    https://doi.org/10.21105/joss.04493

    doi

Pith tools

Reviewed August 16, 2026 · model on record in the stance chip above.