Operationalizing MT Quality Estimation

Track: Multilingual AI | TA2 |
Wednesday, January 27, 2021, 10:30am – 11:00am
Held in: Auditorium
Presenters:
Maribel Rodriguez Molina - Language I/O 
Miklós Urbán - RWS

In this session we will showcase our approach to piloting machine translation (MT) quality estimation (QE) in order to fit this service into production. This approach considers QE scores from a machine learning-based system and uses these scores to provide a more educated guess on the post-editing effort. QE is clearly one of the most awaited features in the MT world. Using QE, we would expect to see a more precise estimation of the actual post-editing effort upfront or anticipate post-editing effort based on how a particular MT engine would perform on content.

Takeaways: Attendees will hear about a methodology for evaluating QE systems; use cases for QE operationalization; and a description of a technology framework supporting MT QE assessment and tracking.