AdiletCodex: A Trilingual (Kazakh-Russian-English) Article-Level Corpus of the Codes and Laws of Kazakhstan

Abstract

AdiletCodex is a full-text, article-level, trilingual (Kazakh + Russian + English) corpus of the currently in-force codes and substantive laws of the Republic of Kazakhstan, parsed from the official legal database adilet.zan.kz. It contains 65,825 article records across 343 documents (24 codes + 319 laws, including the Constitution); 296 documents are present in all three languages (Russian 22,117 + Kazakh 22,099 + English 21,609 article records). Each record carries the full article text, structural hierarchy (part/section/chapter/article), enumerated points, amendment notes, cross-reference links, and rich document metadata (act form, legal sphere, adopting body, registration numbers, adoption/publication dates, status). To our knowledge it is the first article-structured, trilingual full-text corpus of Kazakh legislation. The dataset passed automated quality control (no cross-language duplicate texts, script integrity, article-count alignment). It supports legal information retrieval, cross-lingual legal NLP, statute grounding, legal-translation study, and computational legal research. Source: adilet.zan.kz (official public acts of the Republic of Kazakhstan).Why this corpus. Kazakh legal NLP has lacked a structured, full-text foundation: existing resources are either document-level metadata, Russian-only, or downstream QA sets built on top of statutes they do not themselves release. AdiletCodex fills that gap with the article as the unit of record, the three official language versions side by side, and the structural + amendment + metadata context needed for legal information retrieval, cross-lingual and RAG grounding, translation-fidelity study, and temporal (law-as-of-date) modeling. Related datasets and how AdiletCodex differs. KazakhLawCorpus (Data/MDPI 2026) is document-level metadata only, Kazakh only, no article text; AdiletCodex is complementary and strictly more granular (full text, article-level, trilingual). justicedao/ipfs_kazakhstan_laws_ir (HuggingFace) is article-level but Russian only and undocumented, with no hierarchy/amendments/status/KK/EN; AdiletCodex adds the two other official languages, structure, amendment notes, status, metadata and documentation. Kazakh legal-QA/LoRA sets (e.g. olzhasAl/adilet-legal-qa-kz) are downstream tasks, not a source corpus. The Uzbek legal corpus (Tomaris AI) is a methodological twin for a neighbouring jurisdiction (Uzbek-Latin only, no descriptor). RusLawOD is a Russian national statutory comparator.

Related publications

• A COMPUTATIONAL PIPELINE FOR LEXICAL AND THEMATIC ANALYSIS OF THE CODE OF ADMINISTRATIVE OFFENSES OF THE REPUBLIC OF KAZAKHSTANHerald of Kazakh-British technical university · 2025 · Same author: Bobur Mukhsimbayev
• Anomaly Detection Between Judicial Text-Based DocumentsIEEE International Conference on Application of Information and Communication Technologies · 2020 · Citations: 5 · Same author: Bobur Mukhsimbayev
• Named Entity Recognition Algorithms Comparison For Judicial Text DataIEEE International Conference on Application of Information and Communication Technologies · 2020 · Citations: 3 · Same author: Bobur Mukhsimbayev
• SLeDoC: System for Legal Document ComparisonInternational ACM SIGIR Conference on Research and Development in Information Retrieval · 2026 · Shared topic: Artificial Intelligence in Law