Datasetpublic

VISAI-AI/nitibench-statute

📜 NitiBench-Statute: Thai Legal Corpus for RAG Part of the NitiBench Project This dataset contains the complete corpus of legal sections used in the NitiBench benchmark (CCL and Tax subset). It comprises 5,127 legal sections extracted from 35 Thai legislations (primarily focusing on Corporate and Commercial Law). It is designed to be used as a Context Pool (Knowledge Base) for Retrieval-Augmented Generation (RAG) pipelines. Researchers and developers can load this dataset to… See the full description on the dataset page: https://huggingface.co/datasets/VISAI-AI/nitibench-statute.

sourceHugging Facemitupdated 10mo agoView on Hugging Face

0likes55downloads

Dataset Card

📜 NitiBench-Statute: Thai Legal Corpus for RAG

Part of the [NitiBench Project](https://github.com/vistec-AI/nitibench/)

This dataset contains the complete corpus of legal sections used in the NitiBench benchmark (CCL and Tax subset). It comprises 5,127 legal sections extracted from 35 Thai legislations (primarily focusing on Corporate and Commercial Law).

It is designed to be used as a Context Pool (Knowledge Base) for Retrieval-Augmented Generation (RAG) pipelines. Researchers and developers can load this dataset to populate vector databases or search indices to reproduce NitiBench baselines or evaluate new retrieval strategies.

🚀 Quick Start

Loading the Dataset

You can easily load this dataset using the Hugging Face datasets library:

python

from datasets import load_dataset

# Load the statute corpus
dataset = load_dataset("vistec-AI/nitibench-statute", split="ccl")

# Example: Print the first section
print(dataset[0])

Usage for RAG (Context Pool)

To use this as a retrieval source, you typically iterate through the section_content to create embeddings:

python

documents = []
ids = []

for row in dataset:
    # Use 'section_content' as the text chunk to be indexed
    documents.append(row['section_content'])
    # Use 'law_code' or a combination of name+section as ID
    ids.append(f"row['law_code']-row['section_num']")

# ... Proceed to pass `documents` to your VectorDB or Retriever (e.g., FAISS, ChromaDB, BM25)

📊 Dataset Statistics

—Total Documents: 5,127 sections
—Total Legislations: 35 Legislation (Corporate and Commercial Law)
—Language: Thai

📂 Data Structure

Each row represents a specific section of a law.

Column Name	Type	Description
`law_code`	`str`	Unique identifier for the specific law section (e.g., `ก0123-1B-0001`).
`law_name`	`str`	The official full name of the legislation (e.g., `พระราชบัญญัติการประกอบกิจการพลังงาน พ.ศ. 2550`).
`section_num`	`str`	The specific section number within the Act (e.g., `26`).
`section_content`	`str`	The full text content to be used for retrieval. This includes the law name, section number, and the provision text combined.
`reference`	`list`	A list of cross-references to other laws (if applicable).

Example Data Point

json

{
 "law_code": "ก0123-1B-0001",
 "law_name": "พระราชบัญญัติการประกอบกิจการพลังงาน พ.ศ. 2550",
 "section_num": "26",
 "section_content": "พระราชบัญญัติการประกอบกิจการพลังงาน พ.ศ. 2550 มาตรา 26 ก่อนการออกระเบียบ ข้อบังคับ ประกาศ หรือข้อกำหนดใดของคณะกรรมการซึ่งจะมีผลกระทบต่อบุคคล...",
 "reference": []
}

📝 Citation

If you use this dataset in your research, please cite the NitiBench paper:

bibtex

@inproceedings{akarajaradwong-etal-2025-nitibench,
    title = "{N}iti{B}ench: Benchmarking {LLM} Frameworks on {T}hai Legal Question Answering Capabilities",
    author = "Akarajaradwong, Pawitsapak  and
      Pothavorn, Pirat  and
      Chaksangchaichot, Chompakorn  and
      Tasawong, Panuthep  and
      Nopparatbundit, Thitiwat  and
      Pratai, Keerakiat  and
      Nutanong, Sarana",
    booktitle = "Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing",
    month = nov,
    year = "2025",
    publisher = "Association for Computational Linguistics",
}

@misc{akarajaradwong2025nitibenchcomprehensivestudiesllm,
      title={NitiBench: A Comprehensive Studies of LLM Frameworks Capabilities for Thai Legal Question Answering}, 
      author={Pawitsapak Akarajaradwong and Pirat Pothavorn and Chompakorn Chaksangchaichot and Panuthep Tasawong and Thitiwat Nopparatbundit and Sarana Nutanong},
      year={2025},
      eprint={2502.10868},
      archivePrefix={arXiv},
      primaryClass={cs.CL},
      url={https://arxiv.org/abs/2502.10868}, 
}

⚖️ License

This dataset is provided under the MIT License.