John S. Gage wrote: > > > "aren't really the same but almost so", then they are the same term and > can be tightly mapped to each other. It's when terms are used with a > 50% likelihood to mean something that the term must be discarded, but, > once again, without human intervention. I would say there are > mathematical rules underlying all this. If one day I call a cow a "foo" > and the next I call a dog a "foo", then either "foo" means mammal or it > is meaningless. > I couldn't resist, sorry.... This statement is only true in a highly constrained universe of discourse. Terms can be qualified and changed in meaning by context and by such things as allusion, allegory and metaphor. Now one can argue that the use of language in descriptive medical texts is (or rather should be) just such a highly constrained universe. I submit that not enough study has been done on such texts to warrant such a conclusion, no matter how attractive such a situation presents to known automation techniques. I also warn that the use of a statistical fmeasure to force equivalencies will drive the language to convergence around a set of codes and it will cease to be a human language. Perhaps this is what you want, but I would think about it first.
