thunder-research-group/SNU_Thunder-DeID-annotated_court_judgments

Dataset Card for SNU Thunder-DeID Annotated Court Judgments

Dataset Summary

SNU Thunder-DeID Annotated Court Judgments is a dataset constructed for Named Entity Recognition (NER)-based de-identification of Korean court judgments. The dataset consists of annotated court judgments across multiple criminal case types, aiming to support research and development of NER models for Korean court judgments. It includes sentence-level annotated data from three representative categories of court cases:

• Crime of Violence
• Indecent Act by Compulsion
• Fraud

Supported Tasks

• Named Entity Recognition (NER)
• De-identification / Anonymization of Court Judgments

Languages

• Korean (ko)

---

Dataset Structure

Data Instance

{
  "annotated":"""범죄전력
  피고인은 2020. 9. 25. 청주지방법원에서 사기죄 등으로 징역 2년에 집행유예 3년을 선고받았고, 위 판결은 2021. 9. 17. 확정되었다.
  범죄사실
  피고인은 <<<조합아파트>>>B<<</조합아파트>>>조합아파트(<<<아파트>>>C<<</아파트>>>아파트)의 조합장이자 <<<주식회사>>>D<<</주식회사>>> 주식회사(현 <<<주식회사>>>E<<</주식회사>>> 주식회사)의 실질적인 운영주이다.
  피고인은 2007. 4.경 피해자 <<<내국인이름>>>F<<</내국인이름>>>으로부터 피고인이 시행 중인 <<<조합아파트>>>B<<</조합아파트>>>조합아파트 시행사업자금조로 4억 원을 차용한 후 그 변제약속을 이행하지 아니하였고, 이에 피해자는 2014. 7. 2.경 청주지방법원에서 피고인 및 <<<주식회사>>>D<<</주식회사>>> 주식회사 소유인 청주시 상당구 <<<구아래주소>>>G<<</구아래주소>>> 외 4 필지에 있는 부동산에 대하여 청구금액 607,605,479원에 대한 부동산가압류 결정을 받아 위 부동산에 가압류하게 되었다.
  이에 피고인은 2015. 9. 18.경 청주시 이하 불상지에서 피해자에게 연락하여 "<<<아파트>>>C<<</아파트>>>아파트 부지 전체에 대한 대출을 진행하는데 가압류해제서류를 법무사에게 접수해 주면 대출을 받아서 돈을 갚겠다"고 말하여 피해자가 위 가압류청구금액 중 6억 원을 모두 지급하면 가압류를 해제하여 주겠다고 하고 가압류해제서류를 법무사 사무소에 교부하였다. 그러나 피고인은 같은날 피해자에게 4억 원만 송금한 후 피해자에게 연락하여 "토지 대출을 받을 것이다. 가압류를 해제해주면 이틀 내에 나머지 2억 원을 주겠다."고 거짓말을 하였다. 그러나 사실 피고인은 피해자가 가압류를 해제하도록 하더라도 <<<주식회사>>>D<<</주식회사>>> 주식회사의 70억 원 내지 80억 원 채무로 인해 피해자에게 나머지 2억 원을 지급할 의사나 능력은 없었다.
  피고인은 위와 같이 피해자를 기망하여 이에 속은 피해자로 하여금 위 부동산가압류를 해제하도록 하여 위 가압류의 부담이 없는 부동산을 소유하게 되는 재산상 이익을 취득하였다."""
}

Data Fields

• annotated (string):
  The Annotated Court Judgments of court judgments.
  Entity annotations are provided in-line using markup tokens for NER and de-identification training.

---

Data Splits

Split	Number of Examples	Size (bytes)
crime_of_violence	1500	989,685
indecent_act_by_compulsion	1500	1,207,853
fraud	1500	3,514,767

Total Dataset Size: 5,712,305 bytes
Download Size: 2,277,812 bytes

Usage

The dataset is distributed in three splits:

• crime_of_violence
• indecent_act_by_compulsion
• fraud

Example of loading with datasets library:

from datasets import load_dataset

ds = load_dataset("thunder-research-group/SNU_Thunder-DeID-annotated_court_judgments")
print(ds["fraud"]["annotated"][0])

---

Dataset Creation

The dataset was constructed using publicly available anonymized Korean court judgments.
Since original (unanonymized) judgments are not accessible by law, the dataset was created by applying a structured re-annotation process:

Source Data

• LBox Open: Publicly available anonymized Korean court judgments.

Data Generation Process

• Focus on first-instance criminal court judgments, which contain rich factual descriptions and a variety of direct and quasi-identifiers.
• A hierarchical annotation scheme was developed as part of the SNU Thunder-DeID project (currently unpublished).
• The label taxonomy was designed to support accurate NER-based de-identification and to align with legal and practical requirements for anonymizing Korean court judgments.
• Annotators identified placeholder tokens in anonymized texts and re-labeled them according to the defined scheme.
• The process included four stages: initial labeling, scheme refinement, cross-review for consistency, and final resolution of ambiguities.

Reviewing Process

All data were manually annotated and reviewed by researchers.

To ensure quality:

• No annotator reviewed their own examples.
• Standardized annotation guidelines were applied across annotators.
• The review process included guideline refinement and cross-review for consistency.

---

Additional Information

License

This repository contains original work licensed under the
Creative Commons Attribution-ShareAlike 4.0 International License (CC BY-NC-SA 4.0).

Portions of this repository (including tokenizer vocabulary and/or model weights)
are derived from Meta Llama 3.1 and are subject to the Meta Llama 3.1 Community License. https://github.com/meta-llama/llama-models/blob/main/models/llama3_1/LICENSE

• Creative Commons Attribution-ShareAlike 4.0 License:
  https://creativecommons.org/licenses/by-nc-sa/4.0/

Attribution This dataset is a derivative work based on the
lbox-open dataset,
which is licensed under the CC BY-NC-SA 4.0 License.

Citation Information

If you use SNU Thunder-DeID Annotated Court Judgments in your work, please cite:

@misc{hahm2025thunderdeidaccurateefficientdeidentification,
      title={Thunder-DeID: Accurate and Efficient De-identification Framework for Korean Court Judgments}, 
      author={Sungen Hahm and Heejin Kim and Gyuseong Lee and Hyunji Park and Jaejin Lee},
      year={2025},
      eprint={2506.15266},
      archivePrefix={arXiv},
      primaryClass={cs.CL},
      url={https://arxiv.org/abs/2506.15266}, 
}

Source Data Citation Information

@inproceedings{hwang-etal-2022-data,
    title = "Data-efficient end-to-end Information Extraction for Statistical Legal Analysis",
    author = "Hwang, Wonseok  and
      Eom, Saehee  and
      Lee, Hanuhl  and
      Park, Hai Jin  and
      Seo, Minjoon",
    editor = "Aletras, Nikolaos  and
      Chalkidis, Ilias  and
      Barrett, Leslie  and
      Goanț{\u{a}}, C{\u{a}}t{\u{a}}lina  and
      Preoțiuc-Pietro, Daniel",
    booktitle = "Proceedings of the Natural Legal Language Processing Workshop 2022",
    month = dec,
    year = "2022",
    address = "Abu Dhabi, United Arab Emirates (Hybrid)",
    publisher = "Association for Computational Linguistics",
    url = "https://aclanthology.org/2022.nllp-1.12/",
    doi = "10.18653/v1/2022.nllp-1.12",
    pages = "143--152",
    abstract = "Legal practitioners often face a vast amount of documents. Lawyers, for instance, search for appropriate precedents favorable to their clients, while the number of legal precedents is ever-growing. Although legal search engines can assist finding individual target documents and narrowing down the number of candidates, retrieved information is often presented as unstructured text and users have to examine each document thoroughly which could lead to information overloading. This also makes their statistical analysis challenging. Here, we present an end-to-end information extraction (IE) system for legal documents. By formulating IE as a generation task, our system can be easily applied to various tasks without domain-specific engineering effort. The experimental results of four IE tasks on Korean precedents shows that our IE system can achieve competent scores (-2.3 on average) compared to the rule-based baseline with as few as 50 training examples per task and higher score (+5.4 on average) with 200 examples. Finally, our statistical analysis on two case categories {---} drunk driving and fraud {---} with 35k precedents reveals the resulting structured information from our IE system faithfully reflects the macroscopic features of Korean legal system."
}

Contact Information

If you find something wrong or have question about the dataset, contact [email protected].