Microsoft brings AI-powered translation to end users and developers, whether you’re online or offline
Microsoft Translator has added new capabilities that allow users and developers to get artificial intelligence-powered translations whether or not they have access to the Internet.
The new capabilities allow both end-users and third-party app developers to have the benefit of neural translation technology regardless of whether the device is connected to the cloud or offline.
When using the Microsoft Translator app, end users can now download free AI-powered offline packs. In addition, through the new Translator app local feature preview, Android developers will be able to quickly and easily integrate online and offline AI text translations into their apps.
New AI-powered offline language packs for the Translator apps for Android, iOS, and Amazon Fire
The development comes after two years of work, and it complements Microsoft’s overall effort to make sure developers and users can access AI-powered tools where their data is, whether that’s in the cloud or on a device. That ability, which experts refer to as edge computing, comes as experts are figuring out ways to run powerful AI algorithms without the massive computing power of the cloud.
Microsoft Translator released AI-powered online neural machine translation (NMT) in 2016. Because of the computing power needed to run these high-quality translation models, this capability was only available online. In the latter part of 2017, this capability was made available on specific Android phones equipped with a dedicated AI chip. It allowed their users to get offline translation quality that was on par with the quality of online neural translation.
Building on this initial work, the Translator team was able to further optimize these algorithms, allowing them to run directly on any modern device’s CPU without the need for a dedicated AI chip. The new Translator apps now bring NMT to the edge of the cloud for all Android, iOS, and Amazon Fire devices. Support for Windows devices is coming soon.
These new NMT packs produce higher quality translations, which are up to 23 percent better, and about 50 percent smaller than the previous non-neural offline language packs. These NMT packs are available in Translator’s most popular languages and new NMT languages will be added regularly. For the complete up to date list please check out https://translator.microsoft.com/help/articles/languages.
New Translator local feature preview for Android
For Android developers, the Translator app also now offers a preview of the new local feature, which enables developers to quickly and easily add text translation to any Android app that benefit from translation capabilities.
In addition, thanks to these new NMT offline packs, Android developers can for the first time add offline NMT to their apps, allowing their users to get access to NMT translated content without an Internet connection.
To integrate translation in their app, developers will just need to add some simple code that will use Android’s bound service technology with an AIDL interface to silently call the Translator app. The Translator app will do the rest. If the device is connected to the Internet, the Translator app will retrieve the translation from the Microsoft Translator service on Azure. If Internet connectivity isn’t available, the Microsoft Translator app will use the local NMT offline language packs to deliver this translation back to their app.
The feature is expected to graduate from preview to general availability within 90 days of the preview release.
When the device is online, translations can also leverage customized translation models that match the app and company’s unique terminology.*
Whether the app gets its translations online or offline, the local feature bills the developer’s existing Microsoft Cognitive Services Translator Text API subscription.* There is no need to create a new one and, as if the cloud API is being called directly, requests are not logged for either online or offline translations.
Learn more about how the local feature preview works in our GitHub documentation and sample app.
* During the preview, some features may not be available or may be available at no cost. Please refer to the documentation release notes for more information.
Learn more
