A Dynamic Model of Speech for the Social Sciences