50-60% Reduction in Engine Training Time with New Enhancements to the KantanMT Platform
KantanMT’s Research team in collaboration with ADAPT has delivered an MT optimisation improvement that delivers 60% increase in the build speed required to train its Custom Machine Translation Engines
KantanMT is pleased to announce the release of a platform optimisation technique that improves the performance of KantanMT’s core technology by more than 60%. Now KantanMT clients with large Custom Machine Translation engines can build and retrain their translation engines faster. For example, the build time for a Chinese to English KantanFleet engine with a word count of approx. 60 million source words decreased from 10.5 hours to 4.2 hours representing a 60% reduction in build time.
This benefit to KantanMT’s core technology is available to all members, but it will be clients with larger engines that will see substantial improvements in the speed of their engine build and retraining times. The new optimisation incorporates research into varying translation models similar to the IBM model 2 approach and was jointly developed in a research engagement between ADAPT Centre for Digital Content Technology and KantanLabs. ADAPT Postdoctoral Researcher, Jinhua Du worked on the project with the KantanLabs team at the company’s Dublin office.
“We are delighted with the results of this enhancement to our core technology,” says Tony O’Dowd, Founder and Chief Architect of KantanMT.com. “Even though the KantanMT platform is already extremely fast and scalable, our research team is continually pushing the bar to improve. We are delighted that the outcome of this collaborative project has been fully tested and is now ready to be passed on to our community of users.”
The research collaboration with ADAPT will allow KantanMT to implement, test and evaluate the developments in a suitable staging environment, resulting in an improvement CMT platform. The implementation of this new technology, that will improve the MT user experience to all KantanMT customers.
Researcher, Dr Dimitar Shterionov will publish a research paper on KantanMT’s new engine customisation techniques in early 2017. To know more about KantanMT’s new engine customisation techniques and further research from KantanLabs, email [email protected].
KantanMT.com is a leading SaaS based machine translation platform that enables users to develop and manage Custom Machine Translation engines in the cloud. The innovative technologies offered on the KantanMT.com platform enable users to easily build MT engines in over 760 language combinations, seamlessly integrating into localization workflows and web applications. KantanMT is based in the INVENT Building, DCU Campus, Dublin 9, Ireland.