Využití statistických metod při popisu neverbálních kolokací

This article investigates the statistic function mutual information (MI-score) as applied to various types of lexical combinations, such as multi-word proper names, multi-word terms, idioms and systemic and textual collocations. As a basis for comparison, combinations of the following parts of speech were considered: group A – noun+noun, group B – adjective+adjective, group C – adverb+adverb, group D – adjective+adverb, group E – adjective+adverb, group F – preposition+noun and group G – noun+preposition. MI-scores with values of 15–23 highlight different types of lexical collocations: proper names in group A, terms in group B, idioms in group C, terms in group D and terms in group E. MI-scores with values of 7.0–7.1 relate mostly to systemic or textual collocations. Prepositional collocations are evaluated in the range of the MI-scores 17.5–7.0. The most typical collocations are systemic-textual collocations for group F and multi-word proper names for group G.