The analysis of the used algorithms for the search of unstructured information was carried out, during which it was established that they do not take into account the thematic orientation of the specialized collection of documents. It is proposed to use an annealing simulation algorithm to obtain numerical coefficients used in the information retrieval ranking algorithm system. The results of the standard and modified search algorithm and ranking search results unstructured information. As a result, the value of quality metrics failed improve by an average of 8%, and the average error decreased by 29%

Keywords: information retrieval, ranking, model, quality metrics, relevance, unstructured information