IMLJD: A Computational Dataset for Indian Matrimonial Litigation Analysis

Referee: [Abstract and Dataset Construction] The central empirical claims depend on accurate extraction of quashing petition outcomes (success/failure) from the judgment texts for all 3,613 cases. The manuscript does not report any details on the labeling methodology, inter-annotator agreement, manual validation, or error rates on a sample. Without this, systematic biases in labeling (such as misclassifying conditional orders or withdrawn petitions) could affect the reported rates of 57.6% versus 39.7% and the 19.6-point differential.

Authors: We agree that the manuscript would benefit from greater transparency on how the structured outcome labels were produced. The labels were generated via a hybrid process: automated extraction of key phrases from the final order/disposition sections of each judgment (e.g., 'petition allowed and quashed', 'petition dismissed', 'quashing petition rejected'), followed by manual review of ambiguous or conditional orders by the first author, who has legal training. A random sample of 150 judgments was double-annotated by a second reviewer to assess consistency. We acknowledge that these steps were not described in sufficient detail. In the revised manuscript we will add a new subsection under Dataset Construction that specifies the exact annotation guidelines, provides examples of edge cases (including conditional quashing orders and withdrawn petitions), reports the observed agreement rate, and discusses potential sources of labeling error. This addition will allow readers to evaluate the reliability of the reported success rates. revision: yes

Referee: [Results and Temporal Robustness] While the temporal matching for 2018-2024 is presented to confirm robustness, the paper should provide more specifics on how the matching was performed and whether any adjustments were made for case characteristics beyond the year to ensure the differential is not confounded by other factors.

Authors: The temporal matching consisted of simply restricting both the Supreme Court and Karnataka High Court subsets to the common 2018–2024 window and recomputing the quash rates on this overlapping period; no propensity-score or other covariate matching was applied. We will expand the Results section to state the exact case counts before and after the temporal filter, describe the filtering criterion explicitly, and note that no further adjustments for case characteristics (such as petitioner gender, specific IPC sections, or case complexity) were performed. We will also add a brief discussion acknowledging that residual confounding by unmeasured case features remains possible, while emphasizing that the persistence of the differential after temporal alignment still supports the core observation. If the referee believes additional matching would strengthen the claim, we are open to exploring it with the available metadata in a future extension. revision: partial