Pragmatic evaluations of automated linguistic creativity