Adaptive Neural Machine Translation
The first Adaptive NMT system was created as part of the ModernMT project, led by Translated and co-founded with FBK Research Center, University of Edinburgh, and the European Commission.
ModernMT adds two extra layers of innovation to Neural Machine Translation:
- uses the context of the entire document to predict a better translation of every sentence.
- the model learns in real-time from each correction of the translators.
ModernMT is a top ranking system at WMT18.Learn more
How to access ModernMT
Plugins for translators
ModernMT is available in MateCat as an optional MT engine and in SDL Trados as an extension to the MyMemory plugin.Get Trados Plugin or Get MateCat license
Source code on GitHub
We believe in open-source and open standards. Feel free to access ModernMT code on GitHub and make it even better!Get code
Pre-trained generic and custom models in 31 language pairs, on several billion words of premium data, plus enterprise support.Contact Us
Light Post-Editing: a use case of Adaptive NMT
Light post-editing is a translation technique in which the human translator edits machine translated suggestions, correcting only the most serious errors.
Today, thanks to the improvements made in the neural machine translation field and the quality of the automatic output, post-editing has become a concrete and valid alternative to professional translation in many language pairs, particularly for low-budget projects that require high scalability (e.g. translation of user-generated content such as customer reviews).
To request a light post-editing service, just select the Economy service level among the three options provided in the Instant Quote.
More questions? Get in touch.
Our team is ready to find a solution to your translation needs.Contact us
Hello, I'm Barbara.
How can I help you?
Barbara - Senior Account Manager
Get an instant quote
The easy way to get your documents translated fast.
Buy online in a few clicks.