Machine translation quality prediction scores?

We have too much content and not enough bandwidth to look at every the machine translation, in all languages. On the other hand, 90% of the machine translations are good enough. If Weglot integrates a machine translation quality prediction API, like ModelFront, we could use it to work on only those few machine translations that are below e.g. 10% quality. So we’d get “hybrid” translation - part human, part machine. (Bablic is adding quality prediction for this exact purpose: https://www.reddit.com/r/machinetranslation/comments/13y7507/bablic_a_weglot_competitor_is_adding_machine/) Full disclosure: I’m one of the co-founders of ModelFront (modelfront.com), the main provider of machine translation quality prediction. We actually use Weglot for modelfront.com, and we actually need this to have any hope of translating our content in decent quality in 3 languages, let alone 10 or 100.