Typology of Semantic Relations Extracted by the Instrument RusVectōrēs

Abstract

The question about the types of semantic relations extractable by the Web-service RusVectōrēs is considered. The research urgency is caused by insufficient state of knowledge about linguistic potential of this instrument and distributive semantics in whole. The novelty of the research is seen in the fact that the problem concerning typology of semantic relations extractable on the base of distributional semantics - that is the method that uses automatic processing of big textual data without taking into account linguistic information - is posed for the first time. Classification of semantic relations including eight types of paradigmatic and three types of syntagmatic relations is proposed. It is shown that application of the method under consideration provides the set of semantic associates that consists predominantly of the words that stand in paradigmatic relations with the entry word; and among them hyponyms or co-hyponyms are the most frequent. The question about an effect that the semantic group of the entry word may have on the types of extractable semantic relations is studied. It is shown that the effect can be traced, although it needs additional investigations on more representative data for further specification. The comparative analysis of five semantic groups of the words gave evidence for the abovementioned general trend (about predominance of hyponyms and co-hyponyms) and allowed to produce several specific hypothesis about correlations between the semantic type of an entry word and the types of semantic relations extracted for the word.