PREPOSITIONAL GRAMMAR COMPONENT FOR SYNTACTICAL AND LEXICAL DISAMBIGUATION IN RUSSIAN BASED ON CORPUS STATISTICS

Grammatical description of the sentence generation for the particular language is usually split into special morphological and Syntactical modules applied autonomously: the variance of morphological forms posed into the enumeration of constructional augmentations produces the enormous list of possible expositions of structural complexity paying no attention to the statistical plausibility of a construction in question. The usual method of reducing the complexity score of the Syntactical construction is to put morphological block inside the Syntactical one thus determining item structures strictly possible for this entity, to take into consideration the preferences of item occurrences. The Russian prepositional constructions are the clear case of exuberant variability of the structural complexity in case we are to interpret the meaning of the govenee nouns, its syntactical semantics and the governor element – some full word in a sentence or a predicative or nominal centre. In Russian the ambiguity of interpretation of a prepositional construction is formed by several meanings of primary prepositions plus several noun forms with different senses combined with the preposition plus possible difference of semantic classes implied by the govenee nouns. We construct an ontology for Russian prepositional constructions based on the corpus statistics and propose a sample from the grammatical module aimed at the analysis of the above mentioned structural variables.