Cross-linguistic distributional analyses with frequent frames: the cases of German and Turkish

Abstract

Syntactic categories (e.g., nouns and verbs) are the basic building blocks of grammar. Hence, establishing the syntactic categories of words is an essential step in the acquisition of syntax. As syntactic categories are not directly labeled in the speech directed to children (and also to adults), an important question is how children acquire knowledge about this crucial aspect of their language. While there are no unique markers of syntactic categories, various kinds of correlational cues have been shown to be informative about word categories, such as phonological properties of words (Cassidy & Kelly, 1991; Kelly, 1992; Monaghan, Christiansen, & Chater, 2007; Shi, Morgan, & Allopenna, 1998) and semantic commonalities across words that belong to the same syntactic category (Braine, 1976; Grimshaw, 1981; Pinker, 1984; Schlesinger, 1971). Maratsos and Chalkley (1980) proposed that distributional information in the input—e.g., word co-occurrence patterns—could be a cue for categorizing words. The idea stemmed from the observation that words occurring in the same linguistic context often belong to the same grammatical category (Bloomfield, 1933; Harris, 1951). For example, the fact that cat and mat in (1) are both preceded by the same word would be evidence for a distributional learner that cat and mat belong to the same category.