Classifying idiomatic and literal expressions using vector space representations

We describe an algorithm for automatic classification of idiomatic and literal expressions. Our starting point is that idioms and literal expressions occur in different contexts. Idioms tend to violate cohesive ties in local contexts, while literals are expected to fit in. Our goal is to capture this intuition using a vector representation of words. We propose two approaches: (1) Compute inner product of context word vectors with the vector representing a target expression. Since literal vectors predict well local contexts, their inner product with contexts should be larger than idiomatic ones, thereby telling apart literals from idioms; and (2) Compute literal and idiomatic scatter (covariance) matrices from local contexts in word vector space. Since the scatter matrices represent context distributions, we can then measure the difference between the distributions using the Frobenius norm. We provide experimental results validating the proposed techniques.