Dosyn D. Methodological foundations for the development of intelligent information retrieval systems based on determining the usefulness of knowledge

The thesis solves an important applied scientific problem of design of intelligent information systems for text processing with aim to knowledge discovery. Knowledge is the useful information which utility numerically evaluates as pertinence, that is the importance for solving tasks user has to solve. As an assessment of pertinence, it is proposed to use the increment of the optimal strategy expected utility of an intelligent agent, which is a formal-logical representation of the information needs of a search service customer. The method is based on the use of technologies for syntactic-semantic analysis of text, learning the planning ontology to solve customer tasks and using the Markov decision process (MDP). The developed technology for evaluating the pertinence of text documents, in contrast to the query relevance measure, directly numerically expresses the cost savings of the client of the intelligent information retrieval system for solving his practical problems. The examples of the developed intelligent information retrieval systems, the functioning of which is based on the constructed models, methods and algorithms, are given. Keywords: knowledge discovery, intelligent information retrieval system, ontology, pertinence, optimal strategy, knowledge base, Markov decision process, expected value of perfect information, Bayesian classifier.