The TMS gets a machine translation for each new original segment from the machine translation API, as usual.
The TMS gets a quality prediction for each new machine translated segment from the quality prediction API.
If a machine translated segment is high-quality, like an exact match, the TMS can now just skip human post-editing.
Else, the low-quality machine translated segment is sent to human post-editing as usual.
Translating software docs 4x more efficiently
Translating 3x more product titles and descriptions
Translating customer support emails 7x faster
Citrix and ModelFront panel on quality estimation at SlatorCon
ModelFront research talk on quality prediction at AMTA
Quality estimation on Machine Translate
Hybrid translation on Machine Translate