How to deal with errors in machine translation: Post-editing

Machine Translation output can be incorrect, containing errors that need to be fixed, especially if the text is destined for publication and if it is important that it contains no errors. The task of identifying and fixing these errors is called post-editing (PE). In this chapter, I provide an overview of the PE process, drawing on both academic and industry sources. I explain how PE is generally divided into light and full PE, and describe standard guidelines for each type, homing in on issues that arise in the application of this classification. The chapter also surveys the various types of interface used in PE (including word processing and spreadsheet software, and professional computer-aided translation tools), and modes of interaction (traditional, adaptive or interactive). Finally, concepts and tools used by researchers into PE are described, and particular focus is put on the measurement of temporal, technical and cognitive effort.