Distance phrase reordering for MOSES - User Manual and Code Guide

We describe the implementation of a novel distance phrase reordering (DPR) model for a public domain statistical machine translation (SMT) system - MOSES (http://www.statmt.org/moses/). The model mainly focuses on the application of machine learning (ML) techniques to a specific problem in machine translation: learning the grammatical rules and content dependent changes, which are simplified as phrase reorderings. This document serves two purposes: a user manual for the functions of the DPR model and a code guide for developers.