Predicting gender and age categories in English conversations using lexical, non-lexical, and turn-taking features