Preference Syntagmatics

This paper compares Yorick Wilks’s theory of preference semantics with the evidence of English usage in a large corpus and reports the rationale of a project (in progress) that attaches meanings, not to lexical items, but to contextual patterns, in which each lexical item is normally found. These contexts are based on analysis of a large corpus and stored in a Pattern Dictionary. In addition to other influences, this work is partly inspired by Wilks’s theory of semantic preferences of the 1970s, but there are significant differences. If meanings are attached to words in context instead of in isolation, the formulas needed to express them can express delicate distinctions without being excessively cumbersome. The Pattern Dictionary provides a resource for reducing lexical ambiguity in texts while maintaining interpretative delicacy. The meaning of a word in an unseen document can be estimated by matching its context to one or other of the normal contexts in the Pattern Dictionary, which are themselves explicitly linked to a meaning, called a primary implicature. In the past dozen years corpus analysis has shown with increasing clarity that, although the number of all possible syntagmatic combinations in which each word can participate is vast, and indeed perhaps unlimited, the number of normal syntagmatic combinations is manageably small. Examples are given of verb entries from the Pattern Dictionary