The underlying text analysis algorithms at the moment use English-language similarities to determine groupings, so it's not currently set up to function with other languages.  You can, however, use  to translate the text using Google Translate to English first and then use these variables instead but this is not ideal.