Context-theoretic Semantics for Natural Language: an Overview

Abstract

We present the context-theoretic framework, which provides a set of rules for the nature of composition of meaning based on the philosophy of meaning as context. Principally, in the framework the composition of the meaning of words can be represented as multiplication of their representative vectors, where multiplication is distributive with respect to the vector space. We discuss the applicability of the framework to a range of techniques in natural language processing, including subsequence matching, the lexical entailment model of Dagan et al. (2005), vector-based representations of taxonomies, statistical parsing and the representation of uncertainty in logical semantics.