Arailym-tleubayeva/LegalRAG

Kazakh Legal Text Chunks

Dataset Summary

Kazakh Legal Text Chunks is a processed corpus of official legal texts of the Republic of Kazakhstan, prepared for retrieval-augmented generation (RAG), legal information retrieval, and grounded legal question answering in the Kazakh language.

The dataset contains structure-preserving text chunks derived from publicly available legal and normative documents. It is intended for research and development in:

• legal retrieval,
• legal QA,
• citation-grounded answer generation,
• structure-aware chunking,
• low-resource language NLP.

The chunking, normalization, and metadata preparation were performed by Arailym Tleubayeva.

Source Data

The dataset was prepared from publicly available official legal texts of the Republic of Kazakhstan, including materials available through official legislative sources such as Adilet.

These texts include laws, codes, normative resolutions, and other official legal documents.

Dataset Structure

The current version contains:

• 56,916 chunks
• 5,603 source documents

Columns

• doc_id — document title or identifier
• date — document date as provided in the source
• section_title — section label
• source — source URL
• chunk_id — chunk index within the document
• text — original chunk text
• words — number of words
• char_len — character length
• word_len — token/word length
• normalized_text — normalized version of the chunk for duplicate analysis and retrieval experiments

Creation Process

This dataset is a processed derivative resource created through:

1. collection of publicly available official legal texts;
2. text cleaning and normalization;
3. document chunking into retrieval-friendly segments;
4. metadata preservation for traceability and citation.

The chunking pipeline was developed independently by the dataset author.

Intended Use

This dataset is suitable for:

• legal RAG systems,
• dense and sparse retrieval experiments,
• reranking pipelines,
• legal question answering,
• citation grounding,
• corpus analysis for Kazakh legal NLP.

Possible downstream tasks

• BM25 retrieval
• dense retrieval with sentence embeddings
• hybrid retrieval
• legal passage reranking
• grounded answer generation
• corpus deduplication
• legal chunk classification

Out-of-Scope Use

This dataset is not intended to:

• provide legal advice;
• replace professional legal consultation;
• serve as an official legal information system;
• guarantee up-to-date legal validity without checking the original source.

Users should always verify legal applicability against the official source.

Licensing and Legal Notice

This dataset contains processed chunks of official legal texts of the Republic of Kazakhstan.

Under the Law of the Republic of Kazakhstan on Copyright and Related Rights, official documents of legislative, administrative, and judicial character, as well as their official translations, are not subject to copyright protection.

The present dataset additionally includes independent processing work by the dataset author, including chunking, normalization, formatting, and metadata preparation.

For open reuse and clear attribution of the processing contribution, this dataset is released under:

CC BY 4.0

This means you may share and adapt the dataset with appropriate attribution.

Attribution

If you use this dataset, please cite:

Dataset author

Arailym Tleubayeva
Astana IT University, Kazakhstan

Related publication

This dataset is related to the legal QA / legal AI research direction described in:

D. Rakhimova, A. Turarbek, V. Karyukin, A. Sarsenbayeva, and R. Alieyev,
“Legal AI in low-resource languages: Building and evaluating QA systems for the Kazakh legislation,”
Computers, vol. 14, no. 9, Art. no. 354, 2025.
DOI: 10.3390/computers14090354.

Please also cite the original article if this dataset is used in work related to Kazakh legal QA, legal retrieval, or legal RAG benchmarking. The article is published open access under CC BY 4.0.

Recommended Citation

@dataset{tleubayeva_kazakh_legal_chunks_2026,
  author       = {Arailym Tleubayeva, Birganym Abilkhaiyr and Azamat Orazgali},
  title        = {Kazakh Legal Text Chunks: A Structure-Preserving Chunked Corpus of Official Legal Texts of the Republic of Kazakhstan},
  year         = {2026},
  publisher    = {Hugging Face},
  note         = {Processed legal corpus for retrieval-augmented generation, legal retrieval, and grounded question answering in Kazakh}
}

## Attribution Note
This dataset should be understood as an independently processed chunked corpus built from publicly available official legal texts of the Republic of Kazakhstan. It is related to the research direction presented in Rakhimova et al. (2025),
which should be cited as a relevant prior work and related scholarly source.