Abstract
This paper proposes a method designed to address the problem of information retrieval from unstructured textual data. The search process is conducted in natural language using keyword-based queries, a method commonly employed in modern search engines. We propose a general-purpose intelligent search engine capable of processing multiple key terms based on a combination of contextual language models and combinatorial algorithms related to the subset sum problem, which belongs to the class of NP- complete problems. Thus, a universal multi-key selection machine with improved time and space characteristics is created based on polynomial algorithms. This proposed machine can be utilized to process large datasets in various areas of the law. To achieve the goal, the new polynomial algorithms are based on the subset sum problem, which belongs to the NP-complete class. These algorithms are significantly more efficient in time and space than the existing best polynomial and exponential algorithms. The proposed method can be adapted for processing large volumes of unstructured data in various applied domains, including law, economics, and medicine.