currently public timeline goes unfiltered for processing.
however, the analytics is currently designed only for English language, e.g. the filtering-out of prepositions etc. and also the entire notion of analytics on words (count, correlation, etc.) should be in the context of a language (e.g. "mate" has an entirely different meaning in English and in Argentinean Spanish ;-) ).
I wanted to filter in tweets by language, seeing that Spring Social exposes a property Tweet.languageCode, but experiments with the library showed that it's always null :-( so I couldn't support that.
any ideas?