Microsoft Translator’s highlights of 2020

Posted on December 16, 2020by Microsoft Translator

2020 has been a difficult year around the world, but there has been an occasional bright spot. Here are some of the highlights we saw over the last year.

New languages and dialects

In 2020, we added support for 12 new languages, including seven languages from the Indian subcontinent – Assamese, Gujarati, Kannada, Malayalam, Marathi, Odia, and Punjabi; three languages from Central Asia – Dari, Kazakh, Pashto; one from Southwest Asia – Kurdish (including two dialects: Central and Northern Kurdish); and one from Europe – Irish Gaelic. There are now 74 languages available in Microsoft Translator for text translation.

This year we also added new dialects for existing translation languages: European Portuguese and Canadian French. Previously, Portuguese and French were offered as single language options. Now, you can specifically select either Brazilian or European Portuguese and either French or Canadian French for better translations to your intended dialect.

The details:

Five languages of India: Microsoft adds five languages of India to Microsoft Translator
Odia: Odia Language Text Translation Is Now Available in Microsoft Translator
Assamese: Assamese Text Translation Is Here
Dari and Pashto: Translator Adds Dari and Pashto Text Translation
Kazakh: Қазақ тілі, қош келдің! (Welcome, Kazakh!)
Kurdish: Translator Adds Two Kurdish Dialects for Text Translation
Gaelic: Dia daoibh! Tá Gaeilge againn!
European Portuguese: Portuguese: A Tale of Two Dialects
Canadian French: C’est tiguidou ça! Translator adds Canadian French

Custom Translator version 2

Custom Translator version 2 was a huge technological achievement which significantly improved the quality of custom translations. Custom Translator version 2 was rolled out in two phases to provide two key upgrades: data residency and improved translation quality. In August, phase one of the V2 rollout enabled regional data residency in the Asia Pacific, Canada, Europe, and additional US regions. This allowed customers who previously could not use Custom Translator due to corporate data security, data privacy, and regulatory requirements to keep their training data in the region of their choice. Phase one also improved translation quality for customers using dictionary-only models by switching to Translator’s state-of-the-art transformer architecture.

Phase two, which was released in November, extended transformer architecture to deliver a huge gain in translation quality. Customers can now build custom models with all document types (Training, Testing, Tuning, Phrase Dictionary and Sentence Dictionary) using full text documents, like Office documents, PDFs, HTML and plain text.

Learn more:

Auto mode in the Microsoft Translator app

The Microsoft Translator app made it easy to get your translations so that you can concentrate on your conversation, not the phone. With Auto mode, there’s no more need to push the microphone button when it’s your turn to talk—start the conversation by selecting the languages, press the microphone and begin talking. The app will listen for the two languages and translate what you have said after you finish speaking. The other person can start talking right away when you’re finished and the app will translate automatically.

To make conversations even easier, you can also flip the text on the top portion of the phone so it can be read right side up from the other side and if you are listening to the speech output you can slow down the playback speed to make sure you understand everything clearly.

Auto mode is available in the Microsoft Translator app for Android and iOS.

Learn more:

Virtual networks and private links

Virtual Networks (VNet) are a security feature of the Translator service that became generally available this year. You can use VNet to treat Azure the same as you would your own datacenter, giving you an isolated and highly secure environment to run your virtual machines and applications. VNet can be used in conjunction with private links within your own virtual network. You can use these private IP addresses to access your subscription, and define subnets, access control policies, and more.

Learn more: