Back in 2018, Microsoft first unveiled custom translation feature Custom Translator as part of its Microsoft Translator service. This service provides enterprises with a new way to customize neural machine translation systems. These customizations can be applied to text and speech translation workflows. Today, Microsoft announced the release of The 2nd Edition of Custom Translator, which is of higher quality than Version 1 and is able to keep training data in selected regions.
Customers can now use local data centers in Asia Pacific, Canada, Europe, and more to store training data to accommodate calls from different regions.
Deploying dictionary files through a custom translator model instead of using the dynamic dictionary features in the translator API call simply simplifies the translation process and reduces maintenance costs.
Microsoft expects more than 40% of customers of the training dictionary model to immediately reap the quality improvements of the Custom Translator v2 model.