FIELD: data processing.
SUBSTANCE: invention relates to the field of technologies of computational linguistics for the processing of natural-language texts. The technical result is achieved by executing the text analysis procedures at the stages of sentensising, fragmentation, graphemising, and morphologising, based on morphological, syntactic, and semantic analyses of the components thereof – whole sentences, fragments thereof, and graphemes with symbols; at the stage of patternising, extracting semantically and syntactically coherent collocations (patterns) from the text, wherein text fragments are attributed and indexed at the indexing stage, performancing and post-performancing filtration are performed at the filtration stage, the fragments are merged and reindexed at the cogmentation stage, reformatting and testing are performed at the repatterning stage, semantic analysis of the text fragments is performed at the semanticising stage, and the target linguistic objects are extracted at the extraction stage; wherein the indexing, filtration, cogmentation, and repatterning stages are repeated, executing a certain number of iterations until only one semantic component remains in the analysed list, constituting the main linguistic object.
EFFECT: increase in the quality of machine translation of texts of any degree of complexity.
18 cl, 1 dwg
