Multilingual Processors: a Unified Approach to Semantic and Syntactic Knowledge Presentation

摘要

The methods of linguistic knowledge presentation for natural language (NL) processors operating with unrestricted texts in Russian, English and Italian languages are considered. Contrastive study of language systems, problem-oriented texts analysis and computer experiments carried out within the framework of application developments made it possible to work out the unified syntactic-semantic models valid for analytical (English) and synthetic (Russian) languages and employ these for the multilingual processor design. For con-struction of an integral presentation of sentence semantics structure the semantic structures of various types of predicate expressions which constitute sentence proposition were investigated. The knowledge presentation technique is the extended semantic networks (ESN) mechanism which enables us to simulate highly embedded structures such as compound natural language sentences and superphrasal units. Subject and linguistic knowledge processing is performed by means of production rules. The balance between the "rule-based" approach and the "example-based" approach is observed.