Research of our laboratory is dedicated to the development of model-theoretic methods in knowledge engineering. We develop methods of formal representation and integration of knowledge extracted from a variety of natural language texts. We also develop model-theoretic methods for the generation of new knowledge. These studies are based on the application of model-theoretic methods for the development of ontologies and the logic analysis of natural language, and also on theoretical studies of algebraic systems such as enriched Boolean algebras and normed fields. We carry out research on logical formalization of connections between concepts and on inaccurate and incomplete knowledge representation methods, based on the use of Boolean and fuzzy models.
Our Laboratory achieved the following results:
Head of Laboratory:
- We developed a model-theoretic approach to knowledge extraction from natural language texts. This approach is based on formal representation of extracted knowledge in terms of finite subsets of atomic diagrams of algebraic systems;
- We developed methods for the interpretation of Russian speech diverse parts and diverse syntactic connections, in order to automatically generate algebraic signatures;
- We developed automatization methods aimed at constructing atomic diagrams of algebraic systems from texts in Russian. We also developed methods aimed at the integration of knowledge extracted from natural language texts;
- We developed a software system enabling the generation of fragments of atomic diagrams from natural language texts. This software system allows the user to get answers to questions in Russian and under a certain form, based on the knowledge presented in the model;
- We developed a question-answering system in Russian. Such system searches for information on the internet using parameterized queries;
- We developed a question-answering system with answer reliability assessment, based on the construction of a generalized fuzzy model. In this system, questions are formed by templates, which are probabilistic analogues of templates to "if-questions" and "which-questions";
- We developed methods aimed at extracting knowledge from natural language texts in social networks, based on the Speech-Act theory. Using description language linguistic patterns can solve the problem of identification of Russian language phrases containing “directives” speech acts. It was shown that dialogues containing many of these directives are specifically those which lead to the organization of joint-actions in social networks.
Doctor of Physico-Mathematical Sciences, Professor Dmitry Palchunov, email@example.com
Section of General informatics, Department of Information Technology NSU
Sobolev Institute of Mathematics, Siberian Branch of the Russian Academy of Sciences