Towards a Systematic and Human-Informed Paradigm for High-Quality Machine Translation

Since the advent of modern statistical machine translation (SMT), much progress in system performance has been achieved that went hand-in-hand with ever more sophisticated mathematical models and methods. Numerous small improvements have been reported whose lasting effects are hard to judge, especially when they are combined with other newly proposed modifications of the basic models. Often the measured enhancements are hardly visible with the naked eye and two performance advances of the same measured magnitude are difficult to compare in their qualitative effects. We sense a strong need for a paradigm in MT research and development (R&D), that pays more attention to the subject matter, i.e., translation, and that analytically concentrates on the many different challenges for quality translation. The approach we propose utilizes the knowledge and experience of professional translators throughout the entire R&D cycle. It focuses on empirically confirmed quality barriers with the help of standardised error metrics that are supported by a system of interoperable methods and tools and are shared by research and translation business.