Skip to content
TranslaQE

Machine translation · Quality score · Human review

Machine translation that tells you which sentences to trust.

TranslaQE translates with an engine that adapts to your memories and glossaries, scores every sentence and sends only the ones that need it to human review.

Join the waitlistSee how it works

Early access in batches · EU servers

service-agreement.docx · ES → ENSample output

  1. El contrato entra en vigor el 1 de marzo.

    Trusted92

    The agreement enters into force on 1 March.

  2. Las partes podrán rescindirlo con un preaviso de 30 días.

    Trusted88

    The parties may terminate it with 30 days' notice.

  3. La garantía no cubre los daños por uso indebido.

    Review71

    The warranty does not cover damages for undue use.

  4. Se aplicará la tarifa vigente en el momento del pedido.

    Redo42

    It will apply the tariff in force in the moment of the order.

How it works

  1. 01

    Translate

    Paste text or upload a file (.docx, .xliff, .txt). The engine applies your translation memory and glossaries.

  2. 02

    Every sentence comes with a score

    A 0-100 score per sentence and a traffic light: trusted, review or redo. You know where to look before opening the file.

  3. 03

    Review only what matters

    Low-scoring sentences go to human review. Whatever is reviewed goes back to the engine, so the next translation is better.

Two in-house technologies

What makes TranslaQE different is built by us.

Adaptive AI Translation engine

Machine translation that adapts to each customer with their translation memories (TMX), their glossaries and what their translators review.

In-house QE

The per-sentence quality score, calibrated per language pair and per customer. It learns from every human review.

It adapts on three levels

  1. Instant

    Translation memory and glossary applied to every sentence.

  2. Continuous

    Every reviewed sentence enters your memory right away.

  3. Deep

    We train your own model on your TMX files and your reviews.

Coming soon

TranslaQE Mask: anonymize before you translate

Detects the personal data in your documents, replaces it with placeholders, translates and puts it back. The engine never sees names, emails or account numbers.

  • Names, addresses, emails, phone numbers, ID numbers, IBANs, cards and plates.
  • Reversible mode (pseudonymization) to translate and deliver the full document.
  • Irreversible mode (anonymization) for translation memories and training.
  • Available on the web and the API: /v1/anonymize or the anonymize option in /v1/translate.

Quote post-editing by quality score

Like the match analysis in your CAT tool, but by score: you know how many words are trusted and how many need review before you quote.

  • Servers in the European Union.
  • Your memories and reviews never train other customers' models.
  • You can delete your data and your models at any time.
Sample analysis of a 10,000-word project
Trusted6,420 words · 64 %
Review2,580 words · 26 %
Redo1,000 words · 10 %

Join the waitlist

We open the private beta in batches. We will email you when there is a spot, and nothing else.

The waitlist opens very soon.

We are putting the finishing touches to it. Come back in a few days to sign up.