On the Use of Homogenous Sets of Subjects in Deceptive Language Analysis