Adaptation in Machine Translation
Jan Niehues · Repository KITopen (Karlsruhe Institute of Technology) · 2014
Statistical machine translation (SMT) has emerged as the currently most promising approach for machine translation. One limitation to date, however, is that the quality of SMT systems strongly depends on the similarity between the training data and its deployment. This thesis is devoted to adapting MT systems in the scenario of mismatching training data. We develop different approaches to increase performance even though all or some of the training data does not match the system's application.