NP-COMPLETE ALGORITHMS IN LEGALTECH: HYBRID ARCHITECTURE FOR SEMANTIC RETRIEVAL IN KAZAKHSTAN’S LEGAL ACTS

Abstract

With the growing digitalization of legal practice, there is an increasing need for intelligent systems that can efficiently retrieve and interpret regulatory documents. This study presents a hybrid semantic search system for Kazakhstan’s legal acts that integrates multilingual BERT-based retrieval with NP-complete logical verification. The retrieval module identifies semantically relevant legal clauses, while the verification module applies SAT-based algorithms to ensure logical consistency and compliance with legal conditions. The proposed system achieves higher precision and interpretability compared to traditional models such as BM25. Experimental evaluation on a bilingual corpus confirms that combining neural and symbolic reasoning significantly enhances both accuracy and transparency in legal information retrieval.