A Cohesion Graph Based Approach for Unsupervised Recognition of Literal and Non-literal Use of Multiword Expressions

Abstract

We present a graph-based model for representing the lexical cohesion of a discourse. In the graph structure, vertices correspond to the content words of a text and edges connecting pairs of words encode how closely the words are related semantically. We show that such a structure can be used to distinguish literal and non-literal usages of multi-word expressions.