Edited by Raffaele Simone and Francesca Masini
[Current Issues in Linguistic Theory 332] 2014
► pp. 263–282
Word classes play a fundamental role in many parts of linguistics. Yet, when looking at words individually, there is a large ‘class’ of words that does not behave like any other. This is illustrated here by describing the idiosyncratic behaviour of one such word: half. Half is labeled in many different ways: an adjective, a predeterminer, a quantifier, but it does not behave like any of these. For an accurate modeling of the combinatorics of half, and many other words like it, a more lexically driven approach is needed, since merely assigning them to a word class does not do their behaviour justice. This article describes how the behaviour or half, and other non-classifiable words, can be modeled in an extended version of Corpus Pattern Analysis (CPA), a framework that represents the behaviour of words, typically verbs, by generalizing over their actual occurrences in corpus data.