Applying automatically parsed corpora to the study of language variation