Semantic Representations for Multilingual Natural Language Processing

The paper focuses on the problems of research and integral modeling of semantic representations of language structures for the implementation of knowledge extraction from texts and machine translation. The method of presenting language structures and resolving their ambiguity is based on logical linguistic rules and vector spaces. An effective method of displaying the vector of natural language structures in an expanded space of features for the classification of new language objects and structures has been developed. A focal sample of parallel texts of business and scientific documents in Russian, English and French has been formed for various fields of science and technology; an expanded system of new categories has been developed to enhance the expressive power of the original version of the unification grammar. The multilingual parsing procedures have been developed, which take into account the translational transformations.