Imageability, concreteness, and metaphor identification

This image shows a nice example of our metaphor identification procedure.  Each word in  a paragraph is graphed for its imageability and concreteness score, and
a.) highly imageable words
b.) that do not have a close semantic relationship to the topic of the paragraph
are likely to be metaphors.

The core of the imageability and concreteness rankings comes from psycholinguistic research, and we've expanded this core enormously by using WordNet sister terms to approximate these scores for words not contained in the original database.

Incongruous lexical units and metaphor identification

I picked up this book by Gerard Steen et al. about metaphors at the annual LSA meeting. I need to summarize it for the rest of our team. A key idea in their method for finding metaphors is the identification of 'incongruous lexical units' which may be signals of 'cross-domain mapping'. The 'incongruous' idea is related to work by Charteris-Black and his students.