Exploring semantic annotations to measure post-editing quality
Felipe Almeida Costa, Thiago Castro Ferreira, Adriana Silvina Pagano, Wagner Meira · 2021
This chapter analyses whether the quality of a post-edited text can be explained by its conceptual text complexity as well as additional factors such as the semantic category, quality of the corresponding machine-translation output and post-editing effort. It provides a background to the motivation. The chapter discusses and interprets the output of the interpretable model by indicating significant variables and how each variable impacts post-editing quality. It summarizes the work and main findings, followed by the suggestions to further expand it. The source texts, the automatic and post-edited translations and the findings of the analysis are publicly available in the repository. Post-editing is largely acknowledged as a task meant to fix up the output of a machine-translation system, be that carried out by a human or automatically. When human post-editing is used, the whole operation is also referred to as computer-aided human translation as opposed to automatic machine translation.