KanoonGPT/indian-case-laws

--- pretty_name: Indian Case Laws language: - en license: apache-2.0 tags: - law - legal - india - case-law - judgments - legal-tech - text - tabular - datasets --- # Indian Case Laws **Open Indian case-law data for AI, search, and legal research.** This dataset is part of the **KanoonGPT Open Legal Data Initiative** - an effort to make Indian legal data easier to access, trace, and build on for open-source research, legal tech, and production AI systems. KanoonGPT is building structured Indian legal data and data infrastructure for open-source, research, and enterprise AI applications. Learn more at [kanoongpt.in](https://kanoongpt.in). Repository: [KanoonGPT/indian-case-laws](https://huggingface.co/datasets/KanoonGPT/indian-case-laws) ## Coverage at a glance - Timeframe covered: **1950-2026** (rolling; new judgments are added continuously). - Courts covered: **Supreme Court of India + 25 High Courts**. ### Courts currently covered | Code | Court | |---|---| | `SCI` | Supreme Court of India | | `3~22` | High Court of Punjab and Haryana | | `33~10` | Madras High Court | | `36~29` | High Court for State of Telangana | | `28~2` | High Court of Andhra Pradesh | | `10~8` | Patna High Court | | `32~4` | High Court of Kerala | | `8~9` | High Court of Rajasthan | | `21~11` | High Court of Orissa | | `9~13` | Allahabad High Court | | `22~18` | High Court of Chhattisgarh | | `27~1` | Bombay High Court | | `29~3` | High Court of Karnataka | | `23~23` | High Court of Madhya Pradesh | | `7~26` | High Court of Delhi | | `20~7` | High Court of Jharkhand | | `19~16` | Calcutta High Court | | `24~17` | High Court of Gujarat | | `2~5` | High Court of Himachal Pradesh | | `18~6` | Gauhati High Court | | `1~12` | High Court of Jammu and Kashmir | | `5~15` | High Court of Uttarakhand | | `16~20` | High Court of Tripura | | `14~25` | High Court of Manipur | | `17~21` | High Court of Meghalaya | | `11~24` | High Court of Sikkim | ## Why this dataset exists Indian case law is public, but still difficult to work with at scale as it is locked behind CAPTCHAs and PDFs. Court documents are locked behind captcha, are often fragmented, inconsistently structured, and not packaged for modern ML, search, or analytics workflows. KanoonGPT standardizes this data into an **AI-ready, query-friendly, provenance-aware** format so builders can use it for: - legal search and retrieval - citation-aware RAG pipelines - Indian legal NLP benchmarks - case-law analytics - metadata extraction - legal copilots and agents - downstream model training and evaluation ## What this dataset contains This repository focuses on **structured metadata for Indian court judgments**, starting with **Supreme Court** and **High Court** records. The dataset is designed to preserve both: 1. **clean flattened columns** for search, filtering, analytics, and ML pipelines 2. **full normalized source payloads** for auditability and reproducibility Many records also include **public source links** to the underlying JSON and judgment PDF artifacts. ## Provenance This dataset is built from public Indian court judgment data sourced from the **eCourts judgments ecosystem** and mirrored through the AWS Open Data releases for Indian High Court and Indian Supreme Court judgments. Those AWS pages describe the datasets, their scope, and access points. Upstream open-data references: - [Indian High Court Judgments — AWS Open Data](https://registry.opendata.aws/indian-high-court-judgments/) - [Indian Supreme Court Judgments — AWS Open Data](https://registry.opendata.aws/indian-supreme-court-judgments/) KanoonGPT packages, normalizes, and republishes this data in a Hugging Face-friendly format for research and AI use cases. ## Release variants This dataset is being published in three variants: | Variant | What it contains | Status | |---|---|---| | `sample` | Non-partitioned representative subset for quick exploration, demos, schema inspection, and integration tests. Current rule: for each `(source_path_year, court_code)` group, fetch up to 10,000 rows, then randomize and keep 20 rows. | Available | | `structured` | Full flattened metadata for judgments, including parties, citations, court details, dates, provenance, and quality signals. | Rolling release | | `full` | Structured metadata plus judgment text payloads for retrieval, fine-tuning, and text-heavy downstream tasks. | Coming soon | ## Schema overview The current structured release centers on flattened metadata for HC/SC judgments. ### Identity and dataset fields - `id` - `case_metadata_id` - `dataset_source` - `parser_record_id` - `ingestion_split` ### Parties and case caption - `case_title` - `party_petitioner` - `party_respondent` - `party_caption` ### Legal references - `docket_number` - `cnr_number` - `neutral_citation` - `law_report_citation` ### Court and adjudicators - `court_name` - `court_code` - `bench_name` - `presiding_judge` - `coram_members` - `coram_members_text` ### Dates and disposition - `decision_date` - `registration_date` - `citation_year` - `decision_year` - `disposition_text` ### Source provenance and artifacts - `source_relative_path` - `source_path_year` - `source_path_court_code` - `source_path_bench` - `source_filename` - `source_json_s3_url` - `source_pdf_s3_url` - `language_codes` ### Search, parser, and quality fields - `indexable_text` - `headnote_text` - `normalized_record_json` - `parser_json` - `quality_json` - `created_at` - `updated_at` For parquet publishing stability, `normalized_record_json`, `parser_json`, and `quality_json` are serialized as JSON strings in exported files. ## Example record A sample record may include fields such as: - case title and party names - docket number and CNR number - neutral citation and law report citation - court name and bench - coram / adjudicators - decision date and case disposition - provenance URLs pointing to source JSON and judgment PDF - parser diagnostics and quality flags - a full `normalized_record_json` snapshot for traceability This makes the dataset useful both for lightweight metadata workflows and for provenance-sensitive legal AI systems. ## Design principles ### AI-ready, not just archive-ready This dataset is intended for **real downstream usage** — search, ranking, retrieval, analytics, evaluation, and model-building — not just passive storage. ### Structured first Important legal signals such as parties, citations, dates, bench details, and outcomes are flattened into stable columns instead of remaining buried in raw blobs. ### Traceable to source Records preserve source references, source-path derivations, and normalized JSON so users can validate extracted fields against upstream artifacts. ### Honest about quality Legal data at scale is messy. Parser diagnostics and quality signals are included so downstream users can filter, inspect, or review records instead of assuming uniform quality. ## Why KanoonGPT is publishing this We believe Indian legal AI needs **better open infrastructure**. Open-source legal datasets help researchers, startups, and public-interest builders work from a common foundation instead of repeatedly rebuilding the same ingestion layer from scratch. KanoonGPT’s goal is to contribute usable, well-structured legal data that is easier to explore, benchmark, and build on. ## Responsible use - Verify important legal facts against the original court record and judgment PDF before high-stakes use. - Do not treat this dataset as legal advice. - Source data may contain parser noise, missing fields, inconsistent formatting, or upstream errors. - If you are building end-user systems, add your own validation, citations, and human review layers. - Respect privacy, applicable law, and platform terms in downstream applications. ## Acknowledgements This dataset stands on top of the broader Indian open legal data ecosystem. We are grateful to: - [Pradeep Vanga](https://www.linkedin.com/in/pradeepvanga/) and contributors to the open court-data ecosystem - [vanga/indian-high-court-judgments](https://github.com/vanga/indian-high-court-judgments) - [vanga/indian-supreme-court-judgments](https://github.com/vanga/indian-supreme-court-judgments) - AWS and the [Registry of Open Data on AWS](https://registry.opendata.aws/) for sponsoring public hosting and access infrastructure ## About KanoonGPT KanoonGPT is building Indian legal datasets and data infrastructure for open-source, research, and enterprise AI applications. **Coverage areas:** - ⚖️ Case law - 📜 Bare Acts - 🏷️ Legal metadata - 🧩 Structured legal corpora Website: [kanoongpt.in](https://kanoongpt.in) ## Licensing The dataset is released under the **`apache-2.0`**. ## Citation If you use this dataset, please cite the Hugging Face dataset and, where relevant, the upstream open-data sources. ```bibtex @dataset{kanoongpt_indian_case_laws, author = {KanoonGPT}, title = {Indian Case Laws}, year = {2026}, publisher = {Hugging Face}, url = {https://huggingface.co/datasets/KanoonGPT/indian-case-laws}, note = {Company website: https://kanoongpt.in} } ```pretty_name: 印度判例法 language: - en license: apache-2.0 tags: - 法律 - 司法 - 印度 - 判例法 - 判决 - 法律科技 - 文本 - 表格型 - 数据集 --- # 印度判例法 **面向人工智能、检索与法律研究的开源印度判例数据。** 本数据集隶属于**KanoonGPT开源法律数据计划**，旨在简化印度法律数据的获取、溯源流程，并为开源研究、法律科技与商用人工智能系统的开发提供可复用的数据基础。 KanoonGPT正在构建面向开源、科研与企业级人工智能应用的结构化印度法律数据与数据基础设施。更多信息请访问 [kanoongpt.in](https://kanoongpt.in)。 仓库地址：[KanoonGPT/indian-case-laws](https://huggingface.co/datasets/KanoonGPT/indian-case-laws) ## 数据集覆盖范围概览 - 覆盖时间范围：**1950年-2026年**（滚动更新，持续新增判例）。 - 覆盖法院：**印度最高法院与25个高等法院**。 ### 当前覆盖的法院 | 代码 | 法院名称 | |---|---| | `SCI` | 印度最高法院 | | `3~22` | 旁遮普和哈里亚纳邦高等法院 | | `33~10` | 马德拉斯高等法院 | | `36~29` | 特伦甘纳邦高等法院 | | `28~2` | 安得拉邦高等法院 | | `10~8` | 巴特那高等法院 | | `32~4` | 喀拉拉邦高等法院 | | `8~9` | 拉贾斯坦邦高等法院 | | `21~11` | 奥里萨邦高等法院 | | `9~13` | 阿拉哈巴德高等法院 | | `22~18` | 恰蒂斯加尔邦高等法院 | | `27~1` | 孟买高等法院 | | `29~3` | 卡纳塔克邦高等法院 | | `23~23` | 中央邦高等法院 | | `7~26` | 德里高等法院 | | `20~7` | 贾坎德邦高等法院 | | `19~16` | 加尔各答高等法院 | | `24~17` | 古吉拉特邦高等法院 | | `2~5` | 喜马偕尔邦高等法院 | | `18~6` | 高哈蒂高等法院 | | `1~12` | 查谟和克什米尔邦高等法院 | | `5~15` | 北阿坎德邦高等法院 | | `16~20` | 特里普拉邦高等法院 | | `14~25` | 曼尼普尔邦高等法院 | | `17~21` | 梅加拉亚邦高等法院 | | `11~24` | 锡金邦高等法院 | ## 本数据集的创建初衷 印度判例虽属公开信息，但受限于验证码与PDF格式，大规模处理仍存在较高门槛。法院文书常受验证码限制、内容碎片化且结构不统一，未针对现代机器学习、检索与分析工作流进行标准化封装。KanoonGPT将此类数据标准化为**适配人工智能、支持检索查询、可溯源**的格式，使开发者可将其用于： - 法律检索与召回 - 支持引文的检索增强生成（RAG）流水线 - 印度法律自然语言处理基准测试 - 判例分析 - 元数据提取 - 法律智能助手与AI智能体 - 下游模型训练与评估 ## 本数据集包含内容 本仓库聚焦于**印度法院判例的结构化元数据**，首批覆盖**最高法院与高等法院**的记录。 本数据集兼顾两种数据形式： 1. 用于检索、过滤、分析与机器学习工作流的**标准化扁平化列数据** 2. 用于可审计与可复现的**完整归一化源数据负载** 多数记录还附带关联原始JSON与判例PDF文件的**公开源链接**。 ## 数据溯源 本数据集基于**eCourts判例生态系统**公开的印度法院判例数据构建，并通过AWS开源数据发布项目对印度高等法院与最高法院判例数据进行镜像。AWS相关页面详细说明了数据集的范围、内容与访问方式。 上游开源数据参考： - [印度高等法院判例 — AWS开源数据](https://registry.opendata.aws/indian-high-court-judgments/) - [印度最高法院判例 — AWS开源数据](https://registry.opendata.aws/indian-supreme-court-judgments/) KanoonGPT将此类数据进行封装、归一化，并以适配Hugging Face的格式重新发布，以供科研与人工智能场景使用。 ## 发布变体 本数据集提供三种发布变体： | 变体名称 | 包含内容 | 状态 | |---|---|---| | `sample` | 非分区代表性子集，用于快速探索、演示、模式检查与集成测试。当前规则：对每个`(source_path_year, court_code)`分组，最多提取10,000条记录，随后随机保留20条。 | 已上线 | | `structured` | 完整的判例扁平化元数据，包含当事人、引文、法院详情、日期、溯源信息与质量标识。 | 滚动更新中 | | `full` | 结构化元数据加上判例文本负载，用于检索、微调与文本密集型下游任务。 | 即将上线 | ## 模式概览 当前的结构化发布版本聚焦于高等法院与最高法院判例的扁平化元数据。 ### 标识与数据集字段 - `id` - `case_metadata_id` - `dataset_source` - `parser_record_id` - `ingestion_split` ### 当事人与案件案由 - `case_title` - `party_petitioner` - `party_respondent` - `party_caption` ### 法律引用 - `docket_number` - `cnr_number` - `neutral_citation` - `law_report_citation` ### 法院与审判人员 - `court_name` - `court_code` - `bench_name` - `presiding_judge` - `coram_members` - `coram_members_text` ### 日期与判决结果 - `decision_date` - `registration_date` - `citation_year` - `decision_year` - `disposition_text` ### 源溯源与文件工件 - `source_relative_path` - `source_path_year` - `source_path_court_code` - `source_path_bench` - `source_filename` - `source_json_s3_url` - `source_pdf_s3_url` - `language_codes` ### 检索、解析与质量字段 - `indexable_text` - `headnote_text` - `normalized_record_json` - `parser_json` - `quality_json` - `created_at` - `updated_at` 为保证Parquet格式发布的稳定性，导出文件中的`normalized_record_json`、`parser_json`与`quality_json`均以JSON字符串形式序列化存储。 ## 示例记录 示例记录可包含以下字段： - 案件标题与当事人名称 - 案号与CNR编号 - 中性引文与法律报告引文 - 法院名称与审判庭 - 审判组成人员/裁判者 - 判决日期与案件处置结果 - 指向源JSON与判例PDF的溯源URL - 解析诊断信息与质量标记 - 用于溯源的完整`normalized_record_json`快照 这使得本数据集既可用于轻量元数据工作流，也可用于对溯源要求较高的法律人工智能系统。 ## 设计原则 ### 适配人工智能需求，而非仅用于归档 本数据集旨在支持**实际下游应用**——包括检索、排序、召回、分析、评估与模型构建——而非仅作为被动存储的归档数据。 ### 优先结构化 将当事人、引文、日期、审判庭详情与判决结果等关键法律信息扁平化存储为稳定列，而非保留为原始二进制 blob。 ### 可溯源至原始数据源 记录保留源引用、源路径派生信息与归一化JSON，使用户可对照上游工件验证提取字段的正确性。 ### 如实标注数据质量 大规模法律数据往往存在质量参差不齐的问题。本数据集包含解析诊断信息与质量标识，使下游用户可对记录进行过滤、检查与审核，而非假设所有记录质量一致。 ## KanoonGPT发布本数据集的初衷 我们认为印度法律人工智能领域亟需**更完善的开源基础设施**。开源法律数据集可帮助科研人员、初创企业与公益开发者基于统一基础开展工作，避免重复从零构建相同的数据摄入层。KanoonGPT的目标是提供易用、结构良好的法律数据，便于探索、基准测试与二次开发。 ## 合规使用须知 - 高风险使用前，请对照原始法院记录与判例PDF核实重要法律事实。 - 请勿将本数据集视为法律建议。 - 源数据可能存在解析噪声、字段缺失、格式不一致或上游错误。 - 若构建终端用户系统，请添加您自己的验证、引文与人审流程。 - 在下游应用中，请遵守隐私法规、适用法律与平台服务条款。 ## 致谢 本数据集依托于广泛的印度开源法律数据生态系统，我们感谢以下贡献者与机构： - [Pradeep Vanga](https://www.linkedin.com/in/pradeepvanga/) 及其开源法庭数据生态贡献者 - [vanga/indian-high-court-judgments](https://github.com/vanga/indian-high-court-judgments) - [vanga/indian-supreme-court-judgments](https://github.com/vanga/indian-supreme-court-judgments) - AWS与[AWS开源数据注册表](https://registry.opendata.aws/) 对公共托管与访问基础设施的支持 ## 关于KanoonGPT KanoonGPT正在构建面向开源、科研与企业级人工智能应用的印度法律数据集与数据基础设施。 **覆盖领域：** - ⚖️ 判例法 - 📜 成文法条 - 🏷️ 法律元数据 - 🧩 结构化法律语料库 网站：[kanoongpt.in](https://kanoongpt.in) ## 许可协议 本数据集采用**`apache-2.0`**许可协议发布。 ## 引用方式 若使用本数据集，请引用该Hugging Face数据集，若涉及上游开源数据源，请同时引用上游来源。 bibtex @dataset{kanoongpt_indian_case_laws, author = {KanoonGPT}, title = {Indian Case Laws}, year = {2026}, publisher = {Hugging Face}, url = {https://huggingface.co/datasets/KanoonGPT/indian-case-laws}, note = {Company website: https://kanoongpt.in} }