In this thesis, we address the task of learning models for predicting structured outputs, which take as input a tuple of attribute values and produce as output a structured object. In contrast to classification and regression, where the output is a single scalar value, in our case the output is …