Dublin, Ireland, March 12, 2017 --(PR.com
)-- KantanMT is pleased to announce that users will now be able to build and deploy their own Custom Neural Machine Translation (NMT) engines to translate entire documents in any of the language pairs supported by KantanMT. Clients can create the KantanNeural™ engines directly on the KantanMT platform using their own training data simply by selecting Neural MT as the "engine type."
KantanNeural engines are set up to incorporate some of the most powerful KantanMT features, including the seamless combination of Translation Memory (KantanTotalRecall™) and Machine Translation output, automatic post-editing (PEX) and tokenization exceptions.
This new development follows the news of the launch of NMT stock engines in Legal and Automotive domains, which are available within the extensive KantanFleet™ Library. KantanMT clients however, can now use their own training data from any domain to create the NMT engines.
“We are very excited about this new KantanNeural release,” says Tony O’Dowd. “The KantanLabs researchers and the development team at KantanMT have worked very hard over the past few weeks to bring this new development to the KantanMT community. Our users can now build their own NMT engines, and test them with our A/B Testing feature on KantanLQR. During this beta phase, we are offering our NMT engines to clients at no extra cost.”
The NMT engines take longer to build, when compared to SMT engines, and should be factored into the overall project timeline. KantanLabs is currently working on producing engines from English into French, Spanish, Portuguese, Italian and Thai.
To know more about KantanNeural, mail firstname.lastname@example.org.
KantanMT is a Custom Machine Translation platform that provides the most extensive range of customisation and accuracy management features on the market. Used by some of the worlds’ leading brands, KantanMT delivers translations that are superfast, accurate and brand-consistent. KantanMT ensures data-confidentiality with cloud or on-premise deployments, offers self-managed or fully-serviced implementations, and supports any translation volume in over 760 language pairs.