J00-3003 The standard n-gram models for DA discrimination with lexical cues are probably
hide detail