How to deal with contextual synomys

Just read an interesting post on Lucid Imagination's Blog from Nicholas Chase:
How do you deal with contextual synonyms? For example, “bow” and “ribbon” are synonyms — unless you’re also looking for arrows.


That caught my attention and I read the entire post and slide show - highly recommend.

In short - how a company created an 'automated' testing method based on TREC and use it to tune relevance.  I for one would appreciate the release of SQUAT.