This caught my attention. Synonym rings are where there are a set of related text entries (like "base-ball" and "baseball") that are used to expand search parameters. This will increase the likelihood of a textual match. Nifty. But how should an IA model this information?

The term is a synonym "ring" but really there can be more than just a ring can't there (or would that lead to combinitorial explosion if you connect a series of rings?) Still, there should be a good way to model these relationships, perhaps using bi-directed graph forms? Then you can do depth search, breadth search, etc. to avoid combinitorial boom. Hmm....

Also, I suppose you can use simple box and arrows, but how to capture the symantical relationships? By this I mean in the example above the text difference between "base-ball" and "baseball" and the symantic relationship to "ballgame"? In one case simple substitution, the other may lead to a very broad retrieval.

I suspect this will go well in a chapter on model applications to specific tasks.