Khudoyberdiev Khurshed Atokhonovich, Safarov Ghanijon Habibulloevich

Annotation:

The article presents a systematic review and classification of methods and models for the intelligent analysis of legal texts. Particular attention is paid to the prospects of applying the considered methods to the Tajik language, which is characterized by the absence of large annotated legal corpora and specialized language models. It is shown that the most justified approach for building an intelligent legal system in the Tajik language is a combined one that integrates the construction of a structured normative legal corpus, semantic search based on multilingual vector representations, and constrained answer generation with mandatory grounding in normative sources.

Keywords: text mining; natural language processing; large language models; retrieval-augmented generation; legal texts; low-resource languages; Tajik language; semantic search; legal corpus linguistics.