Pith. sign in

REVIEW 4 major objections 5 minor 25 references

LegalSeg: Unlocking the Structure of Indian Legal Judgments Through Rhetorical Role Classification

T0 review · 4 major / 5 minor · reviewed 2026-08-08 · deepseek-v4-flash

Pith's one-line read LegalSeg, the largest annotated dataset of Indian legal judgments for rhetorical role classification, contains 7,120 documents and 1,487,149 sentences labeled with seven roles, and a benchmark on it shows that a hierarchical BiLSTM-CRF…

desk verdict The dataset is a genuine asset, but the paper's headline benchmark claim does not survive contact with its own confusion matrix. read the letter →

arxiv 2502.05836 v1 pith:ACNQYZLI submitted 2025-02-09 cs.CL cs.AIcs.IRcs.LG

classification cs.CLcs.AIcs.IRcs.LG
keywords rhetoricalroleclassificationlegalNLPIndianjudgmentsdiscoursesegmentationhierarchicalBiLSTM-CRFdatasettransformerbaselinesLLMfine-tuning
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 LegalSeg, which it presents as the largest annotated dataset for rhetorical role classification in legal documents: 7,120 Indian Supreme Court and High Court judgments, split into 1,487,149 sentences, each labeled with one of seven roles (Facts, Issue, Arguments of Petitioner, Arguments of Respondent, Reasoning, Decision, None). Using this dataset as a benchmark, the authors compare sentence-level transformers, a hierarchical transformer over InLegalBERT, a graph neural network, and an instruction-tuned LLaMA, among others. Their central finding is that models that see document context and model label transitions—above all a hierarchical BiLSTM-CRF—outperform models that classify sentences independently. If the claims hold, the field gains a roughly twenty-fold larger Indian legal-role corpus and evidence that legal judgments have a recoverable discourse structure that sequence models can exploit.

What carries the argument

The load-bearing mechanism is the hierarchical BiLSTM-CRF: a BiLSTM that reads sentence embeddings (from sent2vec trained on Indian Supreme Court judgments) in document order, topped by a Conditional Random Field layer that scores the sequence of rhetorical-role labels as a whole. The CRF learns transition constraints such as 'an Issue is typically followed by Arguments, then Reasoning, then a Decision,' and the paper credits this explicit sequential modeling for the best benchmark results. The dataset itself, with its seven-role schema and its split into train/validation/test, is the other central object: it is what makes the model comparison possible and is the paper's main claimed contribution.

What would settle it

Take a random sample of 200 LegalSeg documents, have two independent senior legal experts re-annotate every sentence with the same seven-role schema, and compute Cohen's kappa between the new labels and the released labels; a kappa below roughly 0.6 would indicate that the corpus is too noisy to support the paper's model rankings.

Watch

Extended reading notes

Core claim

LegalSeg is the largest annotated corpus for rhetorical role classification in legal documents, comprising 7,120 judgments and 1,487,149 sentences labeled with seven rhetorical roles. On this corpus, the hierarchical BiLSTM-CRF achieves the highest overall performance, with a macro-F1 of 0.77 and an MCC of 0.68, outperforming all transformer-based models, including ToInLegalBERT (F1 0.62) and the InLegalBERT variants, as well as a graph neural network (F1 0.54) and the instruction-tuned RhetoricLLaMA (F1 0.09). This establishes that modeling sentence order and label transitions is more effective than classifying each sentence in isolation. The benchmark also reveals that the None role dominates (48.5% of sentences) and that confusion between Facts and Reasoning and between petitioner and respondent arguments persists across models.

Load-bearing premise

The benchmark rests on the assumption that ten law students applied the seven rhetorical-role labels consistently across 7,120 documents, yet the paper reports no inter-annotator agreement score.

Editorial extensions

If this is right

  • LegalSeg gives researchers a corpus roughly twenty times larger than the previous largest Indian rhetorical-role dataset, enabling training of data-hungry models for legal NLP.
  • The benchmark establishes that sequence-aware models (BiLSTM-CRF) outperform independent sentence classifiers, meaning legal judgments have discourse structure that can be learned and exploited.
  • Adding neighboring-sentence context improves transformer baselines, and training with predicted previous labels matches or beats true labels, suggesting robust sequential training is viable for role labeling.
  • The persistent confusion between Facts and Reasoning, and between petitioner and respondent arguments, together with the 48.5% None class, defines concrete next targets: class-imbalance handling and finer role definitions.
  • RhetoricLLaMA's low F1 (0.09) indicates that open-source instruction-tuned LLMs are not yet competitive for this structured legal labeling task without further domain adaptation.

Reading between the lines

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

  • Because the dataset exposes the discourse skeleton of judgments, it could be reused as a pretraining or distant-supervision signal for other legal tasks the paper does not evaluate, such as summarization, legal question answering, and explainable judgment prediction.
  • The result that predicted labels rival true labels suggests that explicitly modeling label uncertainty during training—for example with a label-noise-aware loss—might push BiLSTM-CRF-type models further; the paper does not test this directly.
  • The BiLSTM-CRF's edge may depend on the corpus's long documents (averaging about 210 sentences); a reader could test whether a long-document transformer with a CRF head narrows the gap on subsets of shorter judgments.
  • A natural companion study would measure inter-annotator agreement on a sample of LegalSeg documents; such a kappa study would put the benchmark's reliability on a quantified footing.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 5 minor

Summary. The paper introduces LegalSeg, a corpus of 7,120 Indian Supreme Court and High Court judgments annotated at the sentence level with seven rhetorical roles, and evaluates a range of classifiers, including a hierarchical BiLSTM-CRF, TransformerOverInLegalBERT, GNNs, role-aware transformers, and an instruction-tuned LLaMA-2 variant. The authors claim that LegalSeg is the largest dataset for rhetorical role classification in legal documents and that the hierarchical BiLSTM-CRF is the best model, leading to the conclusion that sequential and structural context improves rhetorical role classification.

Significance. If the reported results were reliable, the dataset would be a useful contribution roughly an order of magnitude larger than prior Indian rhetorical-role resources, and the systematic comparison of sequential, hierarchical, and sentence-level models would be informative. Strengths of the work include the stated intention to release data and code, the choice of open-source models rather than proprietary APIs, and the breadth of model families considered. However, the paper is not currently reliable: the advertised dataset size is internally inconsistent, the headline metrics in Table 3 cannot be reproduced from the paper's own confusion matrix in Figure 3, and further figure/table mismatches indicate that the problem is not a single typo. The absence of inter-annotator agreement statistics further weakens the dataset claim. I find no circularity in the evaluation itself; the labels are held out from model training and the benchmark comparisons are empirical.

major comments (4)
  1. [Abstract, Section 1, Table 1, Table 2] The abstract, Section 1, and Table 1 state a total of 14,87,149 sentences for 7,120 documents, but the train, validation, and test sentence counts in Table 2 (11,22,507 + 2,93,370 + 1,49,881 = 15,65,758) sum to roughly 78,609 more sentences, a difference larger than the entire test set. This discrepancy is not cosmetic: it affects the per-document average and the claimed scale of the resource, and it must be resolved before the dataset contribution can be assessed.
  2. [Section 7.1, Table 3, Figure 3] Table 3 reports macro precision 0.78, recall 0.77, F1 0.77, accuracy 0.62, and MCC 0.68 for the Hierarchical BiLSTM-CRF model. Recomputing these metrics from the confusion matrix in Figure 3 gives macro precision approximately 0.27, macro recall approximately 0.28, macro F1 approximately 0.27, and accuracy approximately 0.39; the Issue class alone has only 84 correct predictions out of 1,843 test sentences (recall approximately 0.046). The row sums of Figure 3 also disagree with the test-set label counts in Table 2 (for example, None: 58,470 vs 58,500; AoP: 14,507 vs 14,520; Reasoning: 36,678 vs 36,689). This is not a harmless plotting issue: if Figure 3 is accurate, ToInLegalBERT, whose matrix in Figure 6 reproduces Table 3's macro F1 of approximately 0.62, clearly outperforms the BiLSTM-CRF, reversing the paper's central conclusion. The paper's main empirical claim is therefore unsupported by its own artifact.
  3. [Table 3 and Figures 4 and 13] The mismatch is not confined to the BiLSTM-CRF row. For the MTL model, Table 3 reports accuracy 0.41, while the confusion matrix in Figure 4 gives an overall accuracy of 73,471/149,881, which is approximately 0.49. For InLegalBERT(i-1, label_p, i), Figure 13 shows a degenerate model that assigns only two of the seven labels and has an empty Decision row, yet Table 3 reports F1 0.48 and Section 7.3 discusses the predicted-label variant as a meaningful comparison. These additional inconsistencies show that the numerical results in Table 3 cannot be treated as trustworthy unless every row is reconciled with its corresponding confusion matrix.
  4. [Sections 4.3, 4.4, and 7.5] No inter-annotator agreement statistic is reported. The annotation was performed by ten law students with disagreements escalated to senior experts (Sections 4.3 and 4.4), but no kappa, alpha, or agreement rate is given, and Section 7.5 acknowledges persistent confusion between Facts and Reasoning and between Arguments of Petitioner and Arguments of Respondent. Since LegalSeg's value as a benchmark depends on label reliability, the absence of quantified agreement is a significant gap; the Limitations section's assertion that such issues are 'opportunities for future work rather than major limitations' does not address this gap.
minor comments (5)
  1. [Section 4.5] The annotation schema descriptions are brief, and the overlap between Facts and Reasoning that the paper itself acknowledges in Section 7.5 is not addressed by the role definitions; please make the full annotation guidelines and any boundary-case instructions available with the dataset.
  2. [Section 5.3] The MTL loss L = lambda*L_shift + (1-lambda)*L_RR is not accompanied by the value or tuning procedure for lambda, which is needed because the auxiliary label-shift task is a central component of the reported MTL result.
  3. [Figure 2] The pie chart in Figure 2 is difficult to read for the minority classes (Issue at 1.2% and Decision at 2.1%); a bar chart or a table with counts would convey the distribution more clearly.
  4. [Tables 1 and 2] The Indian lakh/crore comma formatting (14,87,149) is inconsistent with the conventional thousands separators used elsewhere and can confuse readers; please use a single formatting convention consistently.
  5. [Section 7.3] The claim that predicted-label variants 'sometimes' outperform true-label variants is based on a single comparison in Table 3; please specify which comparisons support the claim and add significance testing or variance estimates.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the paper is an empirical dataset-and-benchmark contribution with held-out evaluation, not a derivation that reduces to its inputs.

full rationale

The paper's central claims are empirical: it introduces a new annotated corpus and reports model performances on a fixed train/validation/test split. The labels are human annotations held out from model training, so the reported test metrics are not fitted inputs renamed as predictions. The annotation schema is said to follow prior works (Bhattacharya et al. 2019; Kalamkar et al. 2022; Malik et al. 2022), and some of those prior works share authors with the present paper, but the schema is not the target of the paper's prediction claim; the evaluated task is classification into that schema, and the benchmark numbers are produced by the models on held-out data rather than imported from the cited papers. The self-citations to earlier rhetorical-role work (BiLSTM-CRF, MTL, LEGALToBERT) are used as baselines or architectural inspirations, not as evidence that forces the reported outcomes. The apparent contradiction between Table 3 and Figure 3 for Hier_BiLSTM CRF is a serious internal-consistency/correctness problem for the empirical claims, but it is not a circularity: the discrepancy does not show that any quantity was defined in terms of another or that a fitted parameter was relabeled as a prediction. Similarly, the absence of inter-annotator agreement statistics affects label reliability, not circularity. No step in the paper's derivation chain reduces, by construction or by self-citation, to its own inputs.

Assumptions & free parameters 2 free parameters · 4 assumptions · 0 invented entities

This paper makes no mathematical derivation, so the ledger captures dataset and benchmark assumptions. The main debts are to annotation quality with no reported IAA, the inherited 7-role schema, and unreported training hyperparameters that affect the benchmark numbers.

free parameters (2)
  • MTL task-balancing lambda = not reported
    The MTL loss L = lambda*L_shift + (1-lambda)*L_RR in Section 5.3 requires a balancing weight; its value is never given, so the MTL result is not fully reproducible.
  • Class weights for imbalanced loss = not specified
    Sections A.2, A.4, and A.5 say class weights were applied to handle imbalance, but the actual weights are not listed; these weights affect both Role-Aware and InLegalBERT variant results.
assumptions (4)
  • domain assumption The 7-role rhetorical schema is a meaningful and consistent segmentation across Indian Supreme Court and High Court judgments.
    Invoked throughout Section 4.5; the schema is adopted from prior work but no independent validation is provided for its consistency across courts and legal domains.
  • domain assumption The annotations produced by 10 law students are accurate despite no reported inter-annotator agreement metric.
    Sections 4.3 and 4.4 describe training and senior review but do not quantify agreement, so label reliability is assumed.
  • domain assumption Sentence segmentation with spaCy aligns with the rhetorical units that annotators label.
    Section 4.2 uses spaCy for sentence splitting; any mismatch between sentence boundaries and rhetorical boundaries would add noise to all results.
  • domain assumption The random 70-20-10 document split avoids leakage of related judgments across train and test.
    Section 4.2 describes the split but does not report deduplication or stratification by court, year, or case topic.

how reviews work

0 comments
Cite this review

Pith. "Pith review of LegalSeg: Unlocking the Structure of Indian Legal Judgments Through Rhetorical Role Classification." pith.science (2026). https://pith.science/paper/ACNQYZLI

@misc{pith2026250205836,
  author       = {Pith},
  title        = {Pith review of: LegalSeg: Unlocking the Structure of Indian Legal Judgments Through Rhetorical Role Classification},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/ACNQYZLI}},
  note         = {Machine review of arXiv:2502.05836}
}
read the original abstract

In this paper, we address the task of semantic segmentation of legal documents through rhetorical role classification, with a focus on Indian legal judgments. We introduce LegalSeg, the largest annotated dataset for this task, comprising over 7,000 documents and 1.4 million sentences, labeled with 7 rhetorical roles. To benchmark performance, we evaluate multiple state-of-the-art models, including Hierarchical BiLSTM-CRF, TransformerOverInLegalBERT (ToInLegalBERT), Graph Neural Networks (GNNs), and Role-Aware Transformers, alongside an exploratory RhetoricLLaMA, an instruction-tuned large language model. Our results demonstrate that models incorporating broader context, structural relationships, and sequential sentence information outperform those relying solely on sentence-level features. Additionally, we conducted experiments using surrounding context and predicted or actual labels of neighboring sentences to assess their impact on classification accuracy. Despite these advancements, challenges persist in distinguishing between closely related roles and addressing class imbalance. Our work underscores the potential of advanced techniques for improving legal document understanding and sets a strong foundation for future research in legal NLP.

Figures

Figures reproduced from arXiv: 2502.05836 by the authors.

Figure 2
Figure 2. Distribution of Rhetorical Roles within the [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Confusion matrix for rhetorical role classifica [PITH_FULL_IMAGE:figures/full_fig_p009_3.png] view at source ↗
Figure 4
Figure 4. Confusion matrix for rhetorical role classifica [PITH_FULL_IMAGE:figures/full_fig_p013_4.png] view at source ↗
Figures from the paper (5 more)
Figure 6
Figure 6. Figure 6: Confusion matrix for rhetorical role classifica [PITH_FULL_IMAGE:figures/full_fig_p014_6.png]
Figure 7
Figure 7. Figure 7: Confusion matrix for rhetorical role classifica [PITH_FULL_IMAGE:figures/full_fig_p015_7.png]
Figure 9
Figure 9. Figure 9: Confusion matrix for rhetorical role classifi [PITH_FULL_IMAGE:figures/full_fig_p016_9.png]
Figure 10
Figure 10. Figure 10: Confusion matrix for rhetorical role classifi [PITH_FULL_IMAGE:figures/full_fig_p016_10.png]
Figure 13
Figure 13. Figure 13: Confusion matrix for rhetorical role classifi [PITH_FULL_IMAGE:figures/full_fig_p016_13.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

25 extracted references · 18 canonical work pages

  1. [1]

    online" 'onlinestring :=

    ENTRY address archivePrefix author booktitle chapter edition editor eid eprint eprinttype howpublished institution journal key month note number organization pages publisher school series title type volume year doi pubmed url lastchecked label extra.label sort.label short.list INTEGERS output.state before.all mid.sentence after.sentence after.block STRING...

  2. [2]

    write newline

    " write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 global.max substring 't := if while FUNCTION word.in bbl.in capitalize " " * FUNCT...

  3. [3]

    Paheli Bhattacharya, Shounak Paul, Kripabandhu Ghosh, Saptarshi Ghosh, and Adam Wyner. 2019. Identification of rhetorical roles of sentences in indian legal judgments. In Legal Knowledge and Information Systems, pages 3--12. IOS Press

  4. [4]

    Davide Chicco and Giuseppe Jurman. 2020. The advantages of the matthews correlation coefficient (mcc) over f1 score and accuracy in binary classification evaluation. BMC genomics, 21:1--13

  5. [5]

    Prathamesh Kalamkar, Aman Tiwari, Astha Agarwal, Saurabh Karn, Smita Gupta, Vivek Raghavan, and Ashutosh Modi. 2022. https://aclanthology.org/2022.lrec-1.470 Corpus for automatic structuring of legal documents . In Proceedings of the Thirteenth Language Resources and Evaluation Conference, pages 4420--4429, Marseille, France. European Language Resources A...

  6. [6]

    Soumayan Bandhu Majumder and Dipankar Das. 2020. Rhetorical role labelling for legal judgements using roberta. In FIRE (Working Notes), pages 22--25

  7. [7]

    Vijit Malik, Rishabh Sanjay, Shouvik Kumar Guha, Angshuman Hazarika, Shubham Kumar Nigam, Arnab Bhattacharya, and Ashutosh Modi. 2022. https://doi.org/10.18653/v1/2022.nllp-1.13 Semantic segmentation of legal documents via rhetorical roles . In Proceedings of the Natural Legal Language Processing Workshop 2022, pages 153--171, Abu Dhabi, United Arab Emira...

  8. [8]

    Vijit Malik, Rishabh Sanjay, Shubham Kumar Nigam, Kripabandhu Ghosh, Shouvik Kumar Guha, Arnab Bhattacharya, and Ashutosh Modi. 2021. https://doi.org/10.18653/v1/2021.acl-long.313 ILDC for CJPE : I ndian legal documents corpus for court judgment prediction and explanation . In Proceedings of the 59th Annual Meeting of the Association for Computational Lin...

Show all 25 references
  1. [9]

    Gabriele Marino, Daniele Licari, Praveen Bushipaka, Giovanni Comand \'e , Tommaso Cucinotta, et al. 2023. Automatic rhetorical roles classification for legal documents using legal-transformeroverbert. In CEUR WORKSHOP PROCEEDINGS, volume 3441, pages 28--36. CEUR-WS

  2. [10]

    Ashutosh Modi, Prathamesh Kalamkar, Saurabh Karn, Aman Tiwari, Abhinav Joshi, Sai Kiran Tanikella, Shouvik Kumar Guha, Sachin Malhan, and Vivek Raghavan. 2023. https://doi.org/10.18653/v1/2023.semeval-1.318 S em E val-2023 task 6: L egal E val - understanding legal texts . In ...

  3. [11]

    Isar Nejadgholi, Renaud Bougueng, and Samuel Witherspoon. 2017. A semi-supervised training method for semantic search of legal facts in canadian immigration cases. In Legal knowledge and information systems, pages 125--134. IOS Press

  4. [12]

    Shubham Nigam, Anurag Sharma, Danush Khanna, Noel Shallum, Kripabandhu Ghosh, and Arnab Bhattacharya. 2024. https://aclanthology.org/2024.findings-acl.255 Legal judgment reimagined: P red E x and the rise of intelligent AI interpretation in I ndian courts . In Findings of the ...

  5. [13]

    Shubham Kumar Nigam, Aniket Deroy, Noel Shallum, Ayush Kumar Mishra, Anup Roy, Shubham Kumar Mishra, Arnab Bhattacharya, Saptarshi Ghosh, and Kripabandhu Ghosh. 2023 a . Nonet at semeval-2023 task 6: Methodologies for legal evaluation. In Proceedings of the The 17th Internatio...

  6. [14]

    Shubham Kumar Nigam, Navansh Goel, and Arnab Bhattacharya. 2022. nigam@ coliee-22: Legal case retrieval and entailment using cascading of lexical and semantic-based models. In JSAI International Symposium on Artificial Intelligence, pages 96--108. Springer

  7. [15]

    Shubham Kumar Nigam, Shubham Kumar Mishra, Ayush Kumar Mishra, Noel Shallum, and Arnab Bhattacharya. 2023 b . Legal question-answering in the indian context: Efficacy, challenges, and potential of modern ai models. arXiv preprint arXiv:2309.14735

  8. [16]

    Shounak Paul, Arpan Mandal, Pawan Goyal, and Saptarshi Ghosh. 2023. https://arxiv.org/abs/2209.06049 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

  9. [17]

    TYSS Santosh, Apolline Isaia, Shiyu Hong, and Matthias Grabmair. 2024. Hiculr: Hierarchical curriculum learning for rhetorical role labeling of legal documents. arXiv preprint arXiv:2409.18647

  10. [18]

    M Saravanan, Balaraman Ravindran, and S Raman. 2008. Automatic identification of rhetorical roles using conditional random fields for legal document summarization. In Proceedings of the Third International Joint Conference on Natural Language Processing: Volume-I

  11. [19]

    Jarom \' r S avelka and Kevin D Ashley. 2018. Segmenting us court decisions into functional and issue specific parts. In Legal Knowledge and Information Systems, pages 111--120. IOS Press

  12. [20]

    Hugo Touvron, Louis Martin, Kevin Stone, Peter Albert, Amjad Almahairi, Yasmine Babaei, Nikolay Bashlykov, Soumya Batra, Prajjwal Bhargava, Shruti Bhosale, et al. 2023. Llama 2: Open foundation and fine-tuned chat models. arXiv preprint arXiv:2307.09288

  13. [21]

    Shaurya Vats, Atharva Zope, Somsubhra De, Anurag Sharma, Upal Bhattacharya, Shubham Nigam, Shouvik Guha, Koustav Rudra, and Kripabandhu Ghosh. 2023. Llms--the good, the bad or the indispensable?: A use case on legal statute prediction and legal judgment prediction on indian co...

  14. [22]

    Giulia Venturi. 2012. Design and development of temis: a syntactically and semantically annotated corpus of italian legislative texts. In proceedings of the workshop on semantic processing of legal texts (SPLeT 2012), pages 1--12. Citeseer

  15. [23]

    Vern R Walker, Krishnan Pillaipakkamnatt, Alexandra M Davidson, Marysa Linares, and Domenick J Pesce. 2019. Automatic classification of rhetorical roles for sentences: Comparing rule-based scripts with machine learning. ASAIL@ ICAIL, 2385

  16. [24]

    Adam Wyner, Wim Peters, and Daniel Katz. 2013. A case study on legal case annotation. In Legal Knowledge and Information Systems, pages 165--174. IOS Press

  17. [25]

    Adam Z Wyner. 2010. Towards annotating and extracting textual legal case elements. Informatica e Diritto: special issue on legal ontologies and artificial intelligent techniques, 19(1-2):9--18

Pith tools

Reviewed August 8, 2026 · model on record in the stance chip above.