SDL Machine Translation
FAQ for Trados Studio Users
- How is our Machine Translation offer to Trados Studio users changing?
SDL is now offering a FREE SDL Machine Translation subscription, powered by state-of-the-art Neural Machine Translation technology. The new NMT 2.0 models offer significant improvements in quality over previously available Statistical Machine Translation (SMT) engines.
- What do we now offer as standard?
Trados Studio 2015, 2017 and 2019 users will now receive a free 500K characters subscription per month.
The SDL Machine Translation subscription is available for Studio 2019 directly and for Studio 2015 & 2017 via a Free App available on the SDL Appstore
App name: SDL Machine Translation Cloud (formerly BeGlobal4)
For Studio 2019 users the subscription is also available in in the Language Cloud connector / provider page.
- Is the 500K limit for license or per account?
For Freelancer Translators 500K limit is for license, for LSP or Corporation it is Per Account
- What is happening to the existing Language Cloud Machine Translation SMT packages available for Trados Studio?
Existing SMT packages will continue to run alongside the new NMT 2.0 models and will be available until February 2020. After this time, we will only offer the higher quality NMT 2.0 models, part of the SDL Machine Translation subscriptions. More subscriptions will become available.
- Why have we moved from Statistical to Neural Machine Translation? The market shifted to Neural Machine Translation as it has been proven to provide higher accuracy and fluency over prior techniques. SDL has been at the forefront of machine translation developments and with over 20 years of experience in this field. SDL is using the most modern advances in the field of artificial intelligence to help organizations, governments, Language Service providers and Freelance translators translate content better, faster and with market-leading security.
- What will happen to my Adaptive MT engines, can I still use them, what are offering me to compensate?
Customers who have Adaptive MT engines can continue to use these until further notice, we will advise appropriate alternative in the coming months. Our teams are working on updating the adaptation mechanics to extract the most value from the latest NMT advancements.
- What Neural Machine Translation Language Pairs are available today?
SDL is continuously adding language pairs. Those currently available are listed here:
- What happens if I overuse the 500K characters in a monthly period?
The Free SDL Machine Translation subscription is capped at 500K characters. Our sales team would be happy to discuss your particular needs.
- What is the difference between a Machine Translation engine and a Machine Translation Model?
- As part of this package, Trados Studio users have access to Generic NMT models. Models is a better (and new) name that will replace the term “engine”. Engine is obsolete and has a different meaning in other products. While we referred to MT engines for SMT, we refer to NMT models when we talk about NMT.
- How do I access SDL Machine Translation?
You can access SDL NMT via your Language Cloud connector / provider. Equally you can choose to use the free App. Only Studio 2019 can access the SDL Machine Translation subscriptions via the Language Cloud connector / provider.
- I have a package that contains trained or industry engines?
Industry Engines will be deprecated. We have seen great results from the Generic NMT 2.0 models which outperform previous SMT vertical engines.
- What is the difference in post editing NMT to SMT, what should I look out for?
For more information on post editing Neural Machine Translation please refer the SDL Post Editing Certification available
- How do I apply SDL Machine Translation NMT Models?
You can apply SDL Machine Translation models as you did previously for Language Cloud MT
You will see new ‘Generic NMT Models’ as a new resource as below
- Until end of February2020 you can still access your existing SMT baseline engines as shown below
- What's the difference between AdaptiveMT and the new NMT?