Adaptive MT - Trados 2017 new feature

Self-learning Machine Translation

Avoid the frustrating and time-consuming process of correcting the same machine translation errors time and time again with our break-through AdaptiveMT engines. These engines self-learn from your post-edits so that MT output is unique to your style, content and terminology. Now you can trust your own personal MT engine to form part of your translation DNA!  

Experience the speed of the latest innovation in machine translation (MT). AdaptiveMT continuously learns and improves in real time as users post-edit, delivering MT output unique to your style, content and terminology. 

How does it work?

Before the postedit, the engine has some opinion about how to translate the respective phrase.
When the postedit comes in, the adaptive technology will learn the new translation.
When the same phrase comes up again, there is a competition between the engine's original translation and the newly learned translation. Who wins that competition depends on various statistical factors; it's not always the new translation.
However, if the old translation wins, and the posteditor produces the new translation again, the adaptive technology will learn it again; thus, the third time around, the new translation has a much bigger chance of winning over the old one.  

AdaptiveMT is machine translation that starts of as a basis by using SDL Language Cloud which you have today. So it doesn't matter what other resources you have as this works independently of those.

All you would do is correct the machine translation and then two things would happen when you confirm it:

1. The corrected TU would be sent to your Translation Memory as normal
2. The machine translation engine would take note of your correction so it offered the corrected translation next time

Now, you may well ask why this is useful given you already have this in your TM now? The reason it's useful is because the correction would be applied to other similar segments that are not translated yet but would benefit from the correction you have made. The machine translation will start to adapt to the way you are translating with improved translations for segments that you do not find in your TM.

So your TM/Termbases/Dictionaries have nothing to do with this. It's not the same as training an Engine with thousands or TUs that you have translated in the past. This is much more dynamic.  

4. Studio integration and troubleshooting

  1. Activating SDL Language Cloud in Trados 2017
  2. Deactivating SDL Language Cloud Machine Translation - Trados Studio 2017
  3. How to create an adaptive engine in Trados Studio 2017
  4. How to use SDL Language Cloud in SDL Trados Studio 2015
  5. Adaptive MT - Trados 2017 new feature
  6. The Free subscription for SDL Language Cloud in SDL Trados Studio
  7. No translation from Language Cloud MT - Activating Language Cloud MT on your language pair
  8. There is no Language Cloud translation provided in SDL Trados Studio Project
  9. Connecting to the Language Cloud Account from SDL Trados Studio
  10. Application is not configured for LanguageCloudUrlsConfig
  11. Studio 2015 Language Cloud not automatically inserting the translated segment in the target segment
  12. An error has occurred whilst using the translation provider SDL Language Cloud: Connection to Language Cloud failed
  13. Your subscription does not include access to the requested type of benefit
  14. Your current package is not compatible with SDL Trados Studio Language Cloud
  15. Connection to Language Cloud failed
  16. MS Office Excel Language Cloud plug-in Disabled
  17. How to change LC account in SDL Trados Studio
  18. How to pre-translate in SDL Trados Studio  using Language Cloud
  19. Sufficient permission to access the translation provider SDL Language Cloud could not be obtained
  20. Using API Key for access to Language Cloud
  21. The input BCM JSON should not be null. Parameter name: inputBcmJson
  22. Cannot perform runtime binding on a null reference
  23. Look up segments in MT even if a TM match has been found
  24. Studio charges my Language Cloud account for confirmed segments
  25. Usage incurred during pre-translate in Trados Studio 2017

Feedback and Knowledge Base