P06-1109 2000 ) extends DOP with a maximum likelihood reestimation technique based on the expectation-maximization
hide detail