Introducing TranslaQE: machine translation that tells you which sentences to trust

Machine translation already gets most sentences right. The problem is knowing which ones. Today, whoever reviews machine translation has to read everything with the same attention, because nothing tells them where the errors are.

TranslaQE changes that: every sentence comes with a quality score from 0 to 100 and a traffic light.

• Trusted: the sentence can be delivered without review.
• Review: a translator should look at it.
• Redo: better translate it again.

Two in-house technologies

1. TranslaQE Adapt, the adaptive translation engine. It adapts to each customer: it applies their translation memory and glossaries to every sentence, learns from what their translators review and, on request, we train a dedicated model on their memories.
2. TranslaQE Score, the in-house quality score. The per-sentence quality score, calibrated per language pair and per customer, together with checks that catch typical errors: missing numbers, broken tags or glossary terms that were not respected.

Who it is for

Small agencies and freelance translators who want to quote post-editing by score: just like quoting with your CAT tool's match analysis, but knowing how many words are trusted and how many need review.

Your data stays yours

Servers in the European Union, and a customer's memories and reviews never train other customers' models.

We are opening early access in batches. If you want to try it, join the waitlist.