> IMO Note: science is not about your opinion. > unambiguity is essential Is it? We function in the presence of ambiguity all the time. Sometimes, we let t

> IMO Note: science is not about your opinion. > unambiguity is essential Is it? We function in the presence of ambiguity all the time. Sometimes, we let the ambiguity ride. Sometimes, we take a sentence that is (by itself) ambiguous and let the context disambiguate it. Sometimes, we use the ambiguity for some rhetorical purpose and then resolve it later. There are many sources of ambiguity in natural language---it's one of the things that makes NLP so damn hard---and you can't wave that away by saying that everybody is doing it wrong. (Not that generations of prescriptivist grammarians haven't tried.) > Usually modifiers attach to the closest thing they can, as a matter of well-established convention. First of all, attachment ambiguity is a difficult NLP problem specifically because modifiers don't always attach to "the closest thing they can". Second, to the extent that the statement is true (that low, i.e. "closer", attachment has a higher prior probability than high attachment), it is not a matter of "convention" but a fact about the grammar of a language. Third, "not" is special in all sorts of ways (as is negation in general, in many languages, not just English), and as I mentioned before, this specific kind of ambiguity is well-known, and interesting because it seems so strange, and once again you can't just wave the logic stick and declare that it ought not to be so.