Machine translation (MT) is an ever-present topic on translators’ forums and social media, but nevertheless, it’s hard to keep track of all the new tools and solutions in this field. Mats Linder, author of The Trados Studio Manual, has overcome this problem for Studio users by publishing an article on Trados Studio apps/plugins for MT.
Mats has produced a detailed list of some 20 MT solutions, including free and paid plugins that can be downloaded from the SDL AppStore and integrated automated services that can be activated under translation memory provider settings. In the article, Mats also comments on each tool and gives guidance about API keys and confidentiality levels.
What do I use?
I’ve dabbled with a few MT plugins and translation memory providers from Mats’ list, and used Slate Desktop with some success for confidential, domain-specific texts. Another MT tool I draw on from time to time, which Mats doesn’t mention, is GT4T.
Selective MT use
Google Translate for Translators (GT4T), a tool developed by English to Chinese translator Dallas Cao, has two advantages over other apps:
- GT4T works at a segment or even word level, so you can be very selective about what is sent to Google. I reserve GT4T for specific queries instead of performing an automatic look-up on every segment or pretranslating entire files.
- GT4T works in any Windows program, so you can switch from Studio, DVX or MemoQ to Word and use the same shortcut to translate selected text in a specific language pair. (I’ve reassigned the default Ctrl+J shortcut because it clashes with justify in Word.) For CAT tools, there’s also a shortcut to translate the entire current segment or next segment.
Under Language Pair in the menu, you can reach the setup options to choose from a number of MT engines, including Google Translate PBMT (phrase-based MT) and Google Translate NMT (neural MT). Dictionary look-up and glossary features are also available in GT4T.
GT4T used to be free, but it now has a flexible licensing system based on time (1, 6 or 12 months) or character use (150k, 700k or 1200k characters). If you use it selectively, like me, the last option should last for many years.
The fast-moving field of machine translation
When you read any blog post or article about machine translation, don’t forget to check the date it was published. New options constantly appear on the scene, so articles like the one you’re reading right now will be obsolete before the ink dries!
fast-moving indeed! since he wrote it, another one has popped up: https://www.deepl.com/translator (already integrated into CafeTran, even though the API isn’t yet available)
LOL, Michael – that’s the link I slipped into the last sentence, wondering if anyone would be curious enough to open it!
There is an API, most probably leaked out by DeepL people themselves. Google it you will find open source programs built upon it.
Great stuff! Now I’m looking forward to your investigation of Lilt…
Thanks Emma for writing up. Indeed, being selective about what to send is the initial goal I had in mind when I first worked on GT4T. Thanks for reminding me of this.
You can use DeepL with GT4T now. DeepL also provides alternative translations. You may find it useful as you prefer the selective use of GT4T.
For example, if you select ‘Lovely’ and press ctrl+j, you get a popup containing: charmant, agréable, ravissant, beau.