Your MT program

Affordable machine translation engine training, evaluation, and integration

Trained machine translation brings correct terminology and rendering of product names into the best engines on the market. With an engine customized for your content, you can achieve 17% better overall accuracy and 250% better terminology compliance. This shortens edit distance and provides significant savings.

Custom.MT is a new company that implements machine translation for localization groups and translation companies. We prepare and acquire datasets, train engines using Google, Microsoft, Yandex, Systran, or any other technology, and deliver workshops for project managers and linguists on metrics, fair pay compensation schemes.
Request demo

For:

01

Localization team

Train engines with your company’s terminology, product names, and voice. Own  engines and be independent of translation vendors, escape a lock-in. Benefit from MT savings, increase the efficiency of your budget spend by implementing an advantageous post-editing compensation scheme. Outsource TMX maintenance, dataset preparation, and engine re-training operations. Prepare your PMs, linguists, and LSPs for new levels of productivity and efficiency.
02

LSP

Implement post-editing workflows for a larger part of your translation volume. Increase margins and increase profits by $30-40k per million in revenue. Create and brand your own machine translation engines based on the best technology available. Increase leverage from your translation memories by cleaning them and running automated maintenance operations. Anonymize TMX and remove personal data to comply with GDPR.
03

Product group

Implement trained machine translation as a part of your product: multilingual chat and communications, global eCommerce, SEO-friendly knowledge bases, savvy analytics, opinion mining and information monitoring applications.

Workflow

  1. NDA and DPA
  2. Share TMX with 100k - 2,000k segments
  3. Receive dataset analysis and a quality report
  4. TMX preparation
    • AutoQA errors
    • Remove segments unsuitable for training
    • Optional anonymization
    • Optional additional data acquisition
  5. Engine training 
    • Google
    • Microsoft
    • Systran
    • PangeaMT
    • ModernMT
    • Globaleese
    • Cloud Translation
    • Yandex
    • Promt
    • Any other
  6. Engine evaluation
    • Automatic (BLEU, hLEPOR, METEOR)
    • Human (Edit distance, scoring)
  7. Workflow configuration in CAT tools
  8. Integration
    • via native plugins
    • via Inten.to
    • via API
  9. Team training
    • for linguists
    • for project managers
    • for solution architects
  10. Annual support and engine update retraining
 

Why work with us

  • Professional training, dataset preparation, and TMX maintenance
  • Customizable workflow and metrics
    • use our processes
    • or, we can adapt to your methodology
  • Affordable
    • starting at $3000 per engine
Request demo

Technology Partners

Our team

Konstantin Dranch

Co-founder, Researcher

Philipp Vein

Co-founder, CEO

Dr. Nastya Voloshina

Head of Production

Anna Mikhailova

Partnership Manager

Request a presentation

phone linkedin facebook pinterest youtube rss twitter instagram facebook-blank rss-blank linkedin-blank pinterest youtube twitter instagram