Machine translation post-editing in video game localization

The aim of this study is to explore whether free to use machine translation systems can be used to localize video games, instead of opting for crowdsourcing options. In line with this aim, 30 lines related to in-game races and their sub-races were selected from the language file of the video game called Kenshi. These lines were translated using Google Translate and Amazon Translate, which are free-to-use neural machine translation systems, and the raw machine translation outputs obtained were evaluated by three volunteers in accordance with TAUS "adequacy" and "fluency" error annotation criteria. Then, in accordance with TAUS "Human Translation Quality" guidelines, the volunteers post-edited the raw machine translation outputs. In the light of the obtained data, it has been determined that the raw outputs of the aforementioned machine translation systems cannot be used without post-editing. It has been determined that only 32% of Google Translate's raw machine translation output and 23% of Amazon Translate's raw machine translation output can be used without post-editing. However, by making use of machine translation post-editing, it was observed that the translators' word per hour performance doubled at minimum. Thus, it has been concluded that machine translation systems substantially decrease the translation workload. In addition, it has been determined that video game localization competences have a decisive effect on the editing efforts of the people who conduct machine translation post-editing, and especially "creativity" and "familiarity with games culture" competences cause great differences in edit efforts.