Extensions to the SCARF framework

In this paper, we show simple extensions to the scarf framework. They are generalization of existing known techniques in machine learning or with Hidden Markov Models which are derived for the particular case of Segmental Conditional Random Fields (SCRF). The extensions are Empircal Bayes Risk training, and Mixture Modeling.