Use best-fit MT & reduce the need for human review, enabling rapid assessment and deployment of accurate machine translation models. DeMT™ Estimate API seamlessly connects to your content workflows and provides real-time quality scores customized with your data so the results match your organization for accuracy, tone, and authenticity.
Integrated in
Make informed business decisions
Does a particular MT engine produce better output than another? Is a certain language pair consistently better translated through this or that engine? Now you’ll gather data over time and know the answer.
Set up condition-based workflows
Real-time quality prediction puts you in the driver’s seat: you can set a threshold above which segments are good to go, to reduce time and money spent on post-editing.
Benchmark MT engine performance and know when to post-edit
Find out if an MT engine consistently outperforms others for a particular content type or language pair, and route that content to the right engine.
Choose the right MT engine for your specific needs
Scores are generated from state-of-the-art models like sentence embeddings and the COMET framework.
Models can be trained on your data to adapt to your own standards.
Scores are returned in real-time. Handles any volume effectively.
Plans start as low as 20 M. characters per year, with room to grow.
Start with a simple API call and relax while we handle the rest.
Leverage our simple REST API to submit requests and receive scores seamlessly with on-call technical assistance from our experts.
Choose what score works best for you to compare MT engines.
TAUS QE Score
Proprietary AI system that evaluates how close in meaning two sentences are, and returns a score between 0-1. The higher, the better.
COMET Score
Custom Score
To find out how our quality estimation service could fit within your existing workflows and setup, sign up for a free trial in the Estimate API sandbox environment. You can to test the API, obtain test scores and build your own integration (or use an existing integration). A free trial sandbox account gives you access to our generic model, and includes 500,000 characters across 4 language pairs: EN-DE, EN-ES, EN-FR and EN-IT.
Contact us to get access
TAUS QE is a semantics-based quality score, telling you how close in meaning the two segments are. We use TAUS sentence embedding models in order to calculate this similarity score. For our quality estimation, we offer generic and custom models.
Generic QE Model
Our generic QE model is trained on 100+ languages. See the full list here. For our generic model, we offer the TAUS QE or Comet scores. The generic model is available off-the-shelf.
Custom QE Model
TAUS creates custom models based on your unique content. In order to do this, we collaborate closely with you to create a high-quality dataset with which we train the model. It is possible to train a single model to work for various language pairs or domains, however, the best results are obtained by training custom models that are both domain- and language-pair specific.
Learn how MT Quality Estimation can take your business to the next level.
We’ll help you evaluate & select the best MT engine.
We’ll guide you through which score to use.
We’ll help you save in MT post-editing costs up to 80%.
We’ll provide the best training data to customize and optimize your MT engines.