Using automatically annotated corpora in language variation research