Skip to content
Publications

Talk · International Siberian workshops and tutorials on electron devices and materials · June 1, 2026

Multi-Level Classification of Legal Texts by Document Type and Area of Law in Karakalpak

Hamid Khusinov, Lobar Ganieva, Dilnoza Jamolitdinova, Ziyoda Kasimova, Sevarkhon Maripova, Sirojbek Sharipov

Abstract

This article examines the problem of multi-level classification of legal texts in the Karakalpak language. The proposed approach is based on the simultaneous identification of two document characteristics: its type and branch of law. This approach allows for a more complete description of the legal text and makes the classification results more useful for searching, sorting, storing, and subsequently analyzing documents. The relevance of this work lies in the fact that the volume of digital legal materials is constantly growing, while there are still a limited number of ready-made resources and tools for automatic text processing for the Karakalpak language. This study compiled a corpus of legal documents in the Karakalpak language. The majority of the materials were obtained from an open legal database, after which the texts were translated into Karakalpak and further verified by experts. Open documents from local government agencies were also included in the corpus. Two neural network models were used to solve this problem: mBERT for classification by branch of law and XLM-R for classification by document type. Experiments showed that both models are well suited for the automatic processing of Karakalpak legal texts. The obtained results confirm that multi-level classification can be a useful basis for creating electronic archives, legal search systems and other applied solutions for the legal field.

Artificial Intelligence in LawText and Document Classification TechnologiesTopic Modelinglegal documentsKarakalpak languagedocument type classification

More by Dilnoza Jamolitdinova