But, on the whole non-lexical features are all that is needed to accurately determine the language being spoken when collecting data.