Training of Machine Translation engines

Question

I purchased Trados 2017 Professional edition. I would like to try machine translation engines for Japanese-English pair. When I train TRADOS Adaptive machine translation using documents from several sources, how would the order of inputting these sources impact translation outputs? For example, if I have documents such as a style guide, grammar-based instructions, in-domain glossaries, and in-domain TM publically available from a government website. Would translation outputs be same regardless of the order of uploading each source to the engine?

Paul Filkin · Answer

Hi Yumiko Motomura , 
 
At the moment there is no possibility to create an Adaptive Engine for JP -> EN. The only possibilities are these: 
 
EN -> DE, ES, FR, IT, NL 
FR -> EN 
IT -> EN 
NL -> EN 
 
But when we do support JP -> EN the order of your documents will not make any difference. What could make a difference is how you split up the domain material. So you should probably create an adaptive engine for each domain and then group your source material into those domains. At least that makes sense in my mind. 
 
Regards 
 
Paul

Paul Filkin · Answer

Hi Yumiko Motomura , 
 
To train other engines you need to speak to the developer of those engines as this would be something we have no control over at all. So for example, if you were using the Microsoft MT solution then I believe you could train it based on the information here: 
 
 www.microsoft.com/.../hub.aspx 
 
Then you have a couple of options for a plugin that can use this engine here, one free and one paid for: 
 
MT Enhanced Plugin for Trados Studio 
 appstore.sdl.com/.../ 
 
RyS Microsoft Translator 
 appstore.sdl.com/.../ 
 
Hope these are helpful for you. 
 
Regards 
 
Paul

Trados Studio > 4. Machine Translation

Training of Machine Translation engines