adaptivemt self learning machine translation questions answered

AdaptiveMT, a Self-Learning Machine Translation Engine: Your Questions Answered

Update October 9th 2019: This blog refers to adaptive/statistical machine translation, which has now been replaced by SDL’s powerful neural machine translation (NMT) solution. To learn more about NMT please click here.

Since the release of SDL Trados Studio 2017 our new self-learning machine translation, AdaptiveMT, has been creating a real buzz in the industry.  So, in this blog we have put together the most common questions we have received from you over the last few months, to give you all the answers you’re looking for.

Find out more about AdaptiveMT

Which language pairs are currently supported for AdaptiveMT?

AdaptiveMT currently supports EN > DE, FR, ES, IT, NL and FR, IT and NL > EN.

We are introducing new language pairs on a regular basis, so please click here for an up-to-date list of available language pairs.

How long does it take to adapt my engine and when will I see my results?

When you are using AdaptiveMT, your engine is updated when you post-edit the machine translation (MT) output. Each time you make a change to the MT and confirm the segment the engine will learn from it and apply your correction in real time in the next sentence.

Is my content safe and data private with AdaptiveMT?

SDL does not share your data with any 3rd parties. We only use your data to extract statistical information to improve your own AdaptiveMT engine. If you would like more information on our AdaptiveMT privacy policy, please take a look at our terms and conditions.

I already store my translations in my TM, why do I need AdaptiveMT?

AdaptiveMT is designed to be used when you have no translation memory (TM) match. As AdaptiveMT learns from your post-edits to suit your style and terminology, machine translation mistakes won’t need to be corrected every time. The great thing about this is it will significantly reduce your MT post-editing time when you don’t have a TM match.

Do I need an SDL Language Cloud account to use AdaptiveMT?

AdaptiveMT engines are cloud-based, so you will need to create your own SDL LC account if you do not already have one. When you open SDL Trados Studio 2017, the wizard should prompt you to do this. You also have the option to create your SDL LC account here.

How much does AdaptiveMT cost?

With SDL Trados Studio 2017 you will receive 1 free AdaptiveMT engine, with 400,000 characters to use per month. If you would like to purchase additional AdaptiveMT engines, please click here for the pricing of packages available to you.

How is AdaptiveMT different from SDL LC Custom MT Engines?

The goal of AdaptiveMT and SDL LC Custom MT Engines is to improve your machine translation output quality. However, there are some differences.

AdaptiveMT engines learn in real-time from your post edits, meaning your engine is being constantly trained by you, segment by segment. This is compared to SDL LC Custom MT Engines, which are trained at a specific point in time by using a large volume of customer or domain data.

The other difference is AdaptiveMT provides gradual translation quality improvement over time, whereas the translation quality of custom engines is usually significantly noticeable straight away.

If you are a customer who has a large translation memory, looking to use machine translation within the same domain as your TM, we would recommend using custom engines.

Can I share my AdaptiveMT engine?

If you would like to share your AdaptiveMT engine with a client or freelance translator, you can do this through your SDL LC account using an API key.

Please follow these steps to get started:

Log into your SDL LC account and click ‘Integrations’ in the menu on the left-hand side and click ‘New API Key.’

Create a new API key in SDL Language Cloud

You have an unlimited amount of API keys which you can create. You are also free to delete them at any point to revoke access.

For a translator or client to access the AdaptiveMT engine they will need to login via SDL LC in the SDL Trados Studio project settings using the ‘Sign in with API key’ option.

Sign into SDL Language Cloud with the API key

*Please be aware that a translator or client would have full access to your SDL LC account and cannot be limited to a specific AdaptiveMT engine.

With AdaptiveMT you are able to create engines that are personal to you, these engines could be for specific content type, a specific customer or even just language pair, and this should always be considered when sharing an engine with another user.

I hope this blog has answered some of the questions you have about AdaptiveMT. If you would like to discover more about how adaptive machine translation works, please visit our website. If you have any questions about AdaptiveMT, please ask in the SDL Community or contact us on Live Chat.