Employing AI for Source Quality Improvement

Track: Multilingual AI | TA6 |   Advanced |
Wednesday, July 13, 2022, 2:15am – 3:00am
Held in: Pavilion
Konstantin Savenkov - Intento
Host: Ulrich Henes

During machine translation post-editing (MTPE), some segments are edited much more than others. Adding such edits to the machine translation (MT) training data often does not help, as those are not direct translations. In many cases, the editing involves transcreation, adapting the source for specific locales, or improving translatability. In this session, we will discuss finding such segments in translation memories and addressing these issues in the enterprise content creation workflows, either by copywriting guidelines or automating source-quality improvement before the MT. The results stem from a research project conducted by VMware and Intento.

Takeaways: Attendees will learn how to identify problematic segments in translation memories; optimize the content creation process; and build a more robust MT program.