Custom Translator preview is a feature of the Microsoft Translator service. With Custom Translator, enterprises, app developers, and language service providers can build neural translation systems that understand the terminology used in their own business and industry. The customized translation system will then seamlessly integrate into existing applications, workflows and websites.
Custom Translator allows users to customize Microsoft Translator’s advanced neural machine translation when translating text using the Microsoft Translator Text API , and speech translation using the Microsoft Speech services.
Why should I use Custom Translator?
Get Better Translations
Use your previously translated documents (leaflets, webpages, documentation, etc.) to build a translation system that reflects your domain-specific terminology and style, better than a generic translation system. Users can upload TMX, XLIFF, TXT, DOCX, and XLSX documents.
To make data collection and preparation more effective, the system also accepts data that is parallel at the document level but is not yet aligned at the sentence level. If users have access to versions of the same content in multiple languages but in separate documents, Custom Translator will be able to automatically match sentences across documents. The system can also use monolingual data in either or both languages to complement the parallel training data to improve the translations.
Given the appropriate type and amount of training data it is not uncommon to expect gains between 5 and 10, or even more BLEU points on translation quality by using Custom Translator.
Be productive and cost effective
Training and deploying a custom system is easy and does not require any programming skills.
Using the secure Custom Translator portal, users can through an intuitive interface upload training data, train systems, test them right from the portal and deploy them to a production environment with a simple click. The system will then be available for use at scale within a few hours (actual time depends on training data size).
Custom Translator can also be programmatically accessed through a dedicated API. This allows users that create or update training on a regular basis to manage these processes through their own app or webservice.
While in Preview, there is no charge to train or deploy custom models. Once deployed, the cost of using a custom model to translate content is based on the subscriber’s Translator Text API pricing tier.
Pricing for Custom Translator, when in General Availability, will be based on the subscriber’s Translator Text API pricing tier.
- Training a custom system is 1 character per character of training material in source text.
- Hosting a deployed system is 1M characters per month per deployed system.
- The cost of using a deployed customized translation system is the same as the un-customized systems, which is based on the subscriber’s Translator Text API pricing tier.
Securely translate anytime, anywhere on all your apps and services
Translation systems built with Custom Translator are available through the same cloud-based, high performance, highly scalable Microsoft Translator API that powers billions of translations every day. Custom systems can be seamlessly accessed and integrated into any product or business workflow, and on any device, via the Microsoft Translator API or Microsoft Speech services, through standard REST technology.
How do I get started?
1. Ensure you have a Translator text API key
If you don’t have a key already, learn how to sign up
2. Watch the how to video and read the user guide
The video below will give you a quick overview for how to use Custom Translator and the Custom Translator user guide will teach you everything you need to know to start customizing your translations
3. Log into the Custom Translator portal
You can use your Microsoft account or corporate email to sign into the portal
Ask them on Stack Overflow. We monitor these daily!
Translating legacy Statistical Machine Translation
The legacy Microsoft Translator Hub can be used to translate statistical machine translation. Learn more