adding more bounce classification patterns

[lynx]

I've got a boat load of Syntax bounces (naming that Unkown would be more user friendly) and will extract the patterns out to extend the existing table.

Questions:
1. Do you accept localised (non-english) strings too? I have a few from several languages.
2. Any preference for updating "close enough" patterns vs adding new ones?
3. Are there tests for this? From a quick browse of the tree, I didn't see any.
4. I'll have some concrete classification questions, but that's for later.