A true symbiosis between humans and machines
We have world-leading MT technology that adapts in real-time to a user’s data. But if the translated content is being used as raw MT, there’s no feedback loop to learn from mistakes. At Translated, we have an amazing community of highly-engaged linguists that already work with ModernMT, so we thought - why don’t we connect them in the background, so that all users can benefit from feedback and revisions? The result, the future of machine translation: full automation with a Human-in-the-Loop.
Ongoing human revisions in the background
As you use ModernMT, low confidence translations are automatically flagged and sent to professional translators for revision in the background, in real-time, with no intervention needed.
Targeted Quality Estimation
Our adaptive quality estimation uses your data to rank ModernMT output and identify low confidence translations. You determine how much revision you want done per language in an easy-to-use dashboard.
Real-time adaptive machine translation
ModernMT will continuously learn as the human revised segments are fed back to the model in real-time, on an ongoing basis. This leads to constantly improving translations, with a true Human-in-the-Loop.
Tailor your translation experience
Our Human-in-the-Loop solution empowers you with complete control over the translation process. The user-friendly interface allows you to adjust the level of human review for each language and project, giving you transparency on costs and ensuring you get exactly the level of refinement you need to ensure constantly improving translations for your content.
Trusted by Global Leaders
Join the thousands of satisfied customers who rely on ModernMT for smooth and accurate multilingual communication. Don't take our word for it – hear what they have to say.
«ModernMT lets Airbnb take advantage of state-of-the-art Machine Translation embedded within a human-centered translation process.»
Do you have any questions?
It is suitable for all users of ModernMT. It is particularly effective for enterprise users of MT who are using the output of raw MT directly e.g. in real-time scenarios, publishing MT out as-is. In these cases, you get the speed and scale of MT, but it will continue to learn and improve from the corrections by the Human-in-the-Loop.
Human-in-the-Loop (HITL) functionality is built on top of ModernMT. ModernMT, in its standard form, is an adaptive machine translation model. There is no human intervention in the translation process. HITL gives you the option to revise and improve selected output from ModernMT, using our professional translators, so that in can learn and improve on a frequent, regular basis.
The MT model is updated ongoing basis. We send batches for human revision, and update the model, at least a couple of times each day.
This is fully configurable by you, the user, on a language by language basis. For example, in the user interface, you can set some languages to have no revision. In this case it is standard adaptive ModernMT. Otherwise, you can have different configurations for each language, language pair, or other combinations. For example, you could add a set of language pairs to a group (e.g. Tier 1) and set 10% of segments for revision, where you might create another group (Tier) and set 5% of segments for revision.
As often as you want, there is no restriction.
Yes, everything works the same as it always has with ModernMT. With this solution, we are simply adding human revisions where you want, in order to improve the model for you.