Windows 10: Microsoft Custom Translator now in General Availability

Brink · Dec 6, 2018

Custom Translator, now in general availability, significantly improves the quality of your translations by letting you build your own customized neural translation models tuned with your own pre-translated content. Using Custom Translator, you can translate your product names and industry jargon just the way you want.

With Custom Translator, an extension of the Microsoft Translator Text API, part of the Cognitive Services suite of products on Azure, you can build neural translation models that understand the terminology used in your own business and industry. The customized translation model will then seamlessly integrate into existing applications, workflows, and websites.

Custom Translator can be used with Microsoft Translator’s advanced neural machine translation when translating text using the Microsoft Translator Text API and speech translation using the Azure Cognitive Services Speech Service.

Preview customers of Custom Translator have already noted its improvements on translation quality and its usefulness regardless of the amount of pre-translated, bilingual content available.

Alex Yanishevsky, Senior Manager for machine translation at Welocalize, a leading language service provider, remarked, “Using Custom Translator, we’ve seen very good quality in comparison to other engines. It is very flexible. You can make engines just based on dictionaries if you don’t have enough data, and if you do have enough data you can make an engine based on data plus dictionaries. From the standpoint of customization, having that flexibility is really important.”

How it works

Custom Translator is easy to use and does not require a developer once the call to the Translator service has been properly set up in your app’s code. Custom Translator features a simple and intuitive web app that guides you through the 4-step process of customizing a model:

Upload your data

Train a model

Test the model

Deploy the new customized model to be used in your app

View the process in the image below.

For advanced use, there is also the Custom Translator API (preview) to automate the customization into your workflows.

Building and using custom NMT with Translator is quick, easy, and cost effective. By optimizing how training is performed, and how the Translator runtime incorporates the custom training, our team was able to provide a solution for customizing the Translator NMT models with a training cost that is less than 1% of the cost of training a new neural translation model from scratch. This, in turn, enables Microsoft to provide a cost-effective and simple pricing model to our users.

General availability pricing will go into effect on February 1st, 2019.

Get started now

Ensure you have a Translator Text API key
If you don’t have a key already, learn how to sign up.

Log into the Custom Translator portal
You can use your Microsoft account or corporate email to sign into the portal.

Watch the how-to video and read the documentation.

Questions?
Ask them on Stack Overflow. We monitor these daily!

[youtube]2_xfoI9HV9U[/youtube]
Click to expand...

Source: Customize Microsoft Translator's Neural Machine Translation to translate just the way you want: Custom Translator now in General Availability Translator

:)

Brink · Dec 6, 2018

Customized neural machine translation with Microsoft Translator

Released in preview this week at Build 2018, the new Microsoft Translator custom feature lets users customize neural machine translation systems. These customizations can be applied to both text and speech translation workflows.

Microsoft Translator released neural machine translation (NMT) in 2016. NMT provided major advances in translation quality over the then industry-standard statistical machine translation (SMT) technology. Because NMT better captures the context of full sentences before translating them, it provides higher quality, more human-sounding and more fluent translations.

There are limits to generic NMT models, however. Namely, they are not tuned to provide translations that are unique to a specific business or industry. However, it is too compute intensive, and therefore too expensive and slow, to retrain a new neural system from scratch for all languages and for every company or, even worse, for every product division within a company. Additionally, there is generally not enough parallel data available from the user to expect that the custom data alone would be sufficient to train a new good quality and appropriately adapted system.

Customization is a critical component of high quality machine translations. Customization provides additional context to generic translation models so that translations can reflect a company’s industry, tone and unique terminology.

As a simple illustration, the word “check” could mean many different things in English depending on context: a bank check, a check engine light in your car, a check list or box, etc. The same English word will most likely not always translate into the same word in another language. By providing additional context, the new customization feature can better recognize and translate a bank processing “checks” (pieces of paper) versus an auto manufacturer performing “checks” (verifications).

Of course, this is a simple example; generic NMT systems using the context of a sentence are generally sophisticated enough to resolve single word issues such as this. In the context of longer word sequences translation issues may become more apparent.

Simple as the concept of customization may be, putting it into practice is complex.

Customizing Neural Machine Translation models

To translate any word using a neural machine translation system, each word in a sentence is coded along a 500-dimensional vector representing its unique characteristics within a specific language pair (for example, English and Chinese). Based on the language pair used for training and the training data, the neural network will self-define what these dimensions represent. They could encode simple concepts like gender (feminine, masculine, neutral), politeness level (slang, casual, written, formal, and so on), type of word (verb, noun, and so on), but also any other non-obvious characteristics as derived from the training data. Once trained, the system encodes these words into higher dimensional vectors (1,000 dimensions) to embed both the word’s representation and its context with the full sentence. To learn more about how Neural Machine Translation works, please check our How machine translation works page.

Because of the complexity of these systems, an NMT model will have more than 10 million unique parameters per language. Each of these parameters will be tuned during training before the system is deployed for users and developers to use. This training is not only extremely compute-intensive (days of training using multiple dedicated high-power GPUs), but requires amounts of parallel training data that are, in most cases, several orders of magnitude larger than what a typical enterprise will have access to. For this reason, simply retraining a new NMT from scratch for each new system based on customer’s own data is not a technically or economically feasible option.

How we do it—the nuts and bolts of NMT customization

With careful design and research, using a combination of existing neural, hybrid neural and traditional as well as new techniques, we were able to build a customization feature that overcame these limitations to offer customers the possibility of creating translation models unique to their business and industry.

The starting point is available training data from the user’s own industry or company. To customize a system, a corpus of a minimum of 2,000 parallel sentences (human translated sentences in both the source and target language) is necessary. Working in partnership with Microsoft Translator key partners, and analyzing custom systems’ quality on dozens of data sets, we found that the results are always better than generic NMT, with most of systems showing large quality improvements.

Below are a few examples of BLEU score (an industry standard translation quality measure) improvements for various datasets size and content type. The improvement in quality over the default NMT system can vary from a few to more than 10 BLEU points.

Before commencing training, the custom feature first preprocesses the data. For example, if versions of the same content in multiple languages in separate documents (for example, training manuals, web pages, leaflets) are the only content available, the custom feature will be able to automatically match sentences across documents to generate non-duplicated matched parallel training material. Also, if the user has monolingual data in either or both languages, this data can be used to complement the parallel data to improve the translation model.

The first step the system performs is to start from the pre-existing model and to adapt the existing parameters using the new data. Conceptually, the system starts from the local minima of a general domain system (the best generic translation quality) and tunes it to the local minima relative to the custom data (the best translation quality for the user’s data). From a compute perspective, using this approach has tremendous benefits. By training the custom system from an existing one, the training algorithms are starting from parameters that are already close to what they should be and include information about how to translate concepts that may be outside the small amount of data the user provides. Because of this, the computing power necessary to custom train represents less than 1% of the GPU computing power necessary if training were to start from a blank slate model.

To further accelerate training, several variable learning rate techniques were tested, optimized and integrated into the process.

Not only is this approach effective from a cost and duration perspective (a model can be customized in as little as a few hours on the GPU-powered Azure AI cloud), it still significantly improves translations for the given dataset.

“The key to custom training is having a good initializer for the neural network’s parameters and careful, relentless management of the optimizer’s learning rate,” explains Jon Clark, principal scientist on the Microsoft Translator team. “To make a golf analogy, by relying on our team-built systems as initializers, it’s like we’re already on the green and simply have to putt the ball into the hole with a thoughtful stroke.”

To further improve the custom translation quality, in addition to the neural network for translation that is optimized toward the user’s data, the custom feature also builds a language model over both the user’s parallel data and monolingual data. This further helps the system quickly learn and memorize key aspects of the customer’s unique data.

The resulting new model combines the breadth of the generic Translator NMT model with the unique terminology and tone of the customers parallel and monolingual data.

How to use the Microsoft Translator custom feature

The Microsoft Translator custom feature is available now. It can be used to customize neural text translations using the new Microsoft Translator V3 text API. It can also be used to customize speech translation using the new preview of Cognitive Services Speech. Both the V3 Text API and Cognitive Services Speech are available through the Azure portal with paid and free subscriptions.
The new custom portal can also be used to customize speech to text, and text to speech when used with the preview of Cognitive Services Speech. Learn more about these services.

Learn more:

Customization on the Microsoft Translator website

How machine translation works

Click to expand...

Source: Customized neural machine translation with Microsoft Translator - Microsoft Research

Dennis Mop · Dec 6, 2018

Microsoft Translator

Hi,

One way of stopping Microsoft Translator is by uninstalling it. In order for you to uninstall the app, we recommend that you follow the steps provided below:

Press Windows key and type "translator".

Right-click Translator and select Uninstall.

Follow the instructions on the screen.

Let us know if you have other concerns.

Brink · Dec 6, 2018

Microsoft Translator launching Neural Network based translations

Microsoft Translator is now powering all speech translation through state-of-the-art neural networks.
All speech translation apps that use this service, such as Skype Translator and the Microsoft Translator app for mobile devices, are now using neural network technology. Furthermore, the technology is available to all developers and end-users who want to use the Microsoft Translator speech API to integrate the technology into their favorite apps and services.

In addition to the nine languages supported by the Microsoft Translator speech API, namely Arabic, Chinese Mandarin, English, French, German, Italian, Brazilian Portuguese, Russian and Spanish, neural networks also power Japanese text translations. These ten languages together represent more than 80% of the translations performed daily by Microsoft Translator.

Neural network technology has been used for the last few years in many artificial intelligence scenarios, such as speech and image processing. Many of these capabilities are available through Microsoft Cognitive services. Neural networks are making in-roads into the machine translation industry, providing major advances in translation quality over the existing industry-standard Statistical Machine Translation (SMT) technology. Because of how the technology functions, neural networks better capture the context of full sentences before translating them, providing much higher quality and more human-sounding output.

Try and compare neural network translations at https://translator.microsoft.com/neural
Even though Microsoft’s use of neural networks for speech and text translation is still at an early stage, it is producing superior translations to what SMT provides. As with any new technology (we’re in the so-called ascent phase of the s-curve), we know the quality improvements neural networks provide today are only a first step towards future improvements. You can learn more about neural network-powered translation here.

By leveraging the scale and power of Microsoft’s AI supercomputer and the Microsoft Cognitive Toolkit, the team was able to release ten languages simultaneously; additional languages will be supported over time.
The ten languages are in production today, and available for users for translated Skype calls, for Windows desktop and the Skype Preview app for Windows 10, and in the conversation feature of the Microsoft Translator app for iOS and Android.

Anyone can directly try and compare the improved quality of these new neural network models by testing translations at https://translator.microsoft.com/neural.

In addition, for developers and enterprise customers of the Microsoft Translator API, the deployment into production of these new models comes as Microsoft Translator becomes available for test and purchase on the Azure portal. To get started with the Microsoft Translator API using your Azure subscription, click here. If you are already a Microsoft Translator subscriber and want to learn how to move your subscription to Azure, click here. Neural network-powered translation is available for developers using both the speech and text APIs:

All speech API calls are neural network-powered beginning today.

Text API calls using the “generalnn” standard category are powered by neural networks for translations between the above 10 languages. Learn more about standard categories here.

Click to expand...

Source: Microsoft Translator launching Neural Network based translations for all its speech languages Translator

Windows 10: Microsoft Custom Translator now in General Availability

Microsoft Custom Translator now in General Availability

Microsoft Custom Translator now in General Availability

Microsoft Custom Translator now in General Availability

Microsoft Custom Translator now in General Availability - Similar Threads - Microsoft Custom Translator

Approvals in Microsoft Teams Now Generally Available

Microsoft Defender for Endpoint on iOS is now generally available

Tasks in Microsoft Teams is now generally available

Microsoft Lists in Microsoft Teams is now generally available

PowerPoint Live is now generally available

New Gmail now generally available

Customized neural machine translation with Microsoft Translator

Skype Translator Preview Now Available

Bing Custom Search and Bing Search API V7 Now Generally Available