The cooccurrence of linguistic structures

Abstract

The study of cooccurrences, i. e. the analysis of linguistic units that occur together, has had a profound impact on our view of language. Not only has it contributed greatly to the insight that semi-preconstructed phrases and item-specific knowledge are central to how language works, but it has also led to improved dictionaries and teaching materials. Cooccurrences of various linguistic items have been studied under a variety of names, e. g. collocation, colligation or collostruction. While there are well-understood and fully worked out statistical models for the analysis of cooccurrences of pairs of words, no such model exists for cooccurrences of larger linguistic structures. This situation is remedied by the current work. Building on the well-understood 2 × 2 contingency tables and a graph-based representation of linguistic structures, we develop the generalized cooccurrence model, an explicit formal model for the statistical analysis of cooccurrences of arbitrary linguistic structures. Existing methods for the analysis of two-word cooccurrences and for collostructional analysis are shown to be simply special cases of the generalized cooccurrence model.