Automated Extraction of Paradigmatic Relationships from Natural Language Texts on the Basis of the Complex of Heterogeneous Features

摘要

The paper presents the experience of applying a two-stage semantic analysis of texts to extract paradigmatic relationships and concepts of the subject area. An approach to machine understanding of the text based on the joint application of the methods of linguistic and distributive analysis to form a heterogeneous attribute space, including syntactic, statistical, and semantic features of the text, is considered. Examples of the implementation of heuristic and neural-network analyzers are presented, an assessment of their effectiveness is given, and prospects for the development of the approach are indicated.