Microsoft Translator


Microsoft Translator is a multilingual machine translation cloud service provided by Microsoft. Microsoft Translator is integrated across multiple consumer, developer, and enterprise products; including Bing, Microsoft Office, SharePoint, Microsoft Edge, Microsoft Lync, Yammer, Skype Translator, Visual Studio, Internet Explorer, and Microsoft Translator apps for Windows, Windows Phone, iPhone and Apple Watch, and Android phone and Android Wear.
Microsoft Translator also offers text and speech translation through cloud services for businesses. Service for text translation via the Translator Text API ranges from a free tier supporting two million characters per month to paid tiers supporting billions of characters per month. Speech translation via Microsoft Speech services is offered based on time of the audio stream.
The service supports language systems as of . It also supports 11 speech translation systems that currently power the Microsoft Translator live conversation feature, Skype Translator, and Skype for Windows Desktop, and the Microsoft Translator Apps for iOS and Android.

Development

History

The first version of Microsoft's machine translation system was developed between 1999 and 2000 within Microsoft Research. This system was based on semantic predicate-argument structures known as logical forms and was spun from the grammar correction feature developed for Microsoft Word. This system was eventually used to translate the entire Microsoft Knowledge Base into Spanish, French, German, and Japanese.
Microsoft's approach to machine translation, like most of current modern machine translation systems, is "data driven": rather than relying on writing explicit rules to translate natural language, algorithms are trained to understand and interpret translated parallel texts, allowing them to automatically learn how to translate new natural language text. Microsoft's experience with the LF system led directly to a treelet translation system that simplified the LF to dependency trees and eventually to an order template model, significantly improving in speed and enabling the incorporation of new target languages.
The consumer-facing translation site known as Bing Translator was launched in 2007 and provides free text and website translations on the web. Text is translated directly within the Bing Translator webpage while websites are translated through the Bilingual Viewer tools.
In 2011, the service was extended to include numerous Microsoft Translator products through a cloud-based application programming interface, which supports products available to both consumer and enterprise users. An additional speech translation capability was introduced in March 2016.
In May 2018, an update to the API was introduced. This new version offered neural machine translation as the default method of translating. In addition to translation, the new version features transliteration and a bilingual dictionary to look up words to find alternative translations and to view examples in sentences.
Speech translation was integrated into Microsoft Speech services in September 2018, providing end-to-end speech, speech-to-text, and text-to-speech translation.

Translation methodology and research

Microsoft Translator uses machine translation to create instantaneous translations from one natural language to another. This system is based on four distinct areas of computer learning research seen below.
Type of learningImpact on translation
Neural networks
Neural machine translation
Neural networks try to mimic how the brain works to translate between languages. At a high-level, neural network translation works in two stages. First, a first stage models the word that needs to be translated based on the context of this word within the full sentence. Second, the neural network translates this word model, within the context of the sentence, into the other language. Neural machine translation is the default translation method for the Microsoft Translator API.
Syntax-based SMT
Syntax-based translation is based on the idea of translating syntactic units, rather than a word or string of words. Microsoft has used syntax-based SMT to translate much of its computer-related texts from English into multiple target languages. Ongoing research in this area has produced improvements in word inflections and word ordering.
Phrase-based SMTIn phrase-based SMT, the machine learns correspondence between languages from parallel text without the aid of linguist knowledge. This produces better translations in less time than other systems.
Bitext word alignmentSMT systems rely on existing translated data to learn how to automatically translate from one language to another. To train the systems, identifying word correspondences is crucial. Microsoft has developed work in both discriminative and generative approaches to word alignment, resulting in faster algorithms and higher quality and translations.
Language modelingLanguage modeling uses n-gram models to construct comprehensible translations in the target language. This ensures that the output translation is fluent and readable.

Accuracy

The quality of Microsoft Translator’s machine translation outputs are evaluated using a method called the BLEU score.
BLEU is an algorithm for evaluating the quality of text which has been machine-translated from one natural language to another. Quality is considered to be the correspondence between a machine's output and that of a human. BLEU was one of the first metrics to achieve a high correlation with human judgments of quality, and remains one of the most popular automated and inexpensive metrics.
Because machine translation is based on statistical algorithms rather than human translators, the automatic translations it produces are not always entirely accurate. Microsoft Translator has introduced various feedback features, such as the Collaborative Translations Framework, into its products to allow users to suggest alternative translations. These alternative translations are then integrated into the Microsoft Translator algorithms to improve future translations.
In November 2016, Microsoft Translator introduced translation using deep neural networks in nine of its highest traffic languages, including all of its speech languages and Japanese. Neural networks provide better translation than industry standard statistical machine learning.

Core products

Microsoft Translator is a cloud-based API that is integrated into numerous Microsoft products and services. The Translator API can be used on its own and can be customized for use in a pre-publishing or post-publishing environment. The API, which is available through subscription, is free for lower translation volumes, and is charged according to a tiered payment system for volumes exceeding two million characters per month. The remaining core products are available for free.

Microsoft Translator cloud translation

The Microsoft Translator is a cloud-based automatic translation service that can be used to build applications, websites, and tools requiring multi-language support.
Custom Translator is a feature of the Microsoft Translator services that allows enterprises, app developers, and language service providers to build neural translation systems that understand the terminology used in their own business and industry. The customized translations can then be delivered into existing applications, workflows and websites using a normal call to the Microsoft Translator API. Custom Translator can be used when translating text with the Microsoft Translator Text API and when translating speech with Microsoft Speech services.

Live feature

A personal universal translator that enables up to 500 people to have live, multi-device, multi-language, in person translated conversations. This feature is currently free and available in the Microsoft Translator apps and from the browser at: http://translator.microsoft.com.

Microsoft Translator Hub

The Microsoft Translator Hub allows enterprises and language service providers to build their own translation systems that understand business- and industry-specific terminology. The Hub can also be used in conjunction with the CTF, allowing administrators to approve CTF results and add them directly to the Hub. The Microsoft Translator Hub is only available for statistical machine translation and cannot be used with the newest version of the Microsoft Translator API.
The Hub has also been used for language preservation, allowing communities to create their own language translation systems for language and cultural preservation. The Hub has been used to create translation systems for languages such as Hmong, Mayan, Nepali, and Welsh.

Multilingual App Toolkit (MAT)

The Multilingual App Toolkit is an integrated Visual Studio tool, which allows developers to streamline localization workflows of their Windows, Windows Phone and desktop apps. MAT improves localization of file management, translation support, and editing tools.

Microsoft Translator web app

Microsoft Translator is a user facing translation portal provided by Microsoft as part of its Bing services to translate texts or entire web pages into different languages. All translation pairs are powered by the Microsoft Translator, a statistical machine translation platform and web service, developed by Microsoft Research, as its backend translation software. Two transliteration pairs and Chinese ) are provided by Microsoft's Windows International team.
As of , Bing Translator offers translations in different language systems.
's :fr:Accueil|main page.
Bing Translator can translate phrases entered by the user or acquire a link to a web page and translate its entirely. When translating an entire web page, or when the user selects "Translate this page" in Bing search results, the Bilingual Viewer is shown, which allows users to browse the original web page text and translation in parallel, supported by synchronized highlights, scrolling, and navigation. Four Bilingual Viewer layouts are available:
Bing Translator integrates with several other Microsoft products. The following is a table of products into which Bing Translator is integrated or may be integrated:
Integrates intoMean of integration
Bing Instant AnswersAlready integrated
Internet ExplorerAn Accelerator for Internet Explorer 8 or higher

Supported products

Through its core product offerings, Microsoft Translator supports the translation features of many Microsoft products at the consumer and enterprise levels. These products fall broadly into three categories — communication products, Microsoft Office, and apps.

Communication

Collaborative Translations Framework (CTF)

The Collaborative Translations Framework is an extension of the Microsoft Translator API that allows post-publishing improvement of translated text. By using the CTF, readers have the ability to suggest alternative translations to those provided by the API, or vote on previously offered alternatives. This information is then delivered to the API to improve future translations.

Translator Web Widget

The Translator Web Widget is a translation tool that can be added to web pages by pasting a predefined snippet of JavaScript code into the page. The web widget is offered for free by Microsoft, and supports both pre-publishing customized translations using the Translator Hub, and post-publishing improvements using the Collaborative Translations Framework.

Supported languages

As of , Microsoft Translator supports 73 different language systems. The list of supported languages is available at the Microsoft Translator website and can also be retrieved programmatically through the cloud services.
  1. Afrikaans
  2. Arabic
  3. Bengali
  4. Bosnian
  5. Bulgarian
  6. Cantonese
  7. Catalan
  8. Chinese
  9. Chinese
  10. Croatian
  11. Czech
  12. Danish
  13. Dutch
  14. English
  15. Estonian
  16. Fijian
  17. Filipino
  18. Finnish
  19. French
  20. Galician
  21. Georgian
  22. German
  23. Greek
  24. Haitian Creole
  25. Hebrew
  26. Hindi
  27. Hmong Daw
  28. Hungarian
  29. Icelandic
  30. Indonesian
  31. Irish
  32. Italian
  33. Japanese
  34. Klingon
  35. Korean
  36. Latvian
  37. Lithuanian
  38. Malagasy
  39. Malay
  40. Malayalam
  41. Maltese
  42. Maori
  43. Norwegian
  44. Persian
  45. Polish
  46. Portuguese
  47. Portuguese
  48. Punjabi
  49. Querétaro Otomi
  50. Romanian
  51. Russian
  52. Samoan
  53. Serbian
  54. Serbian
  55. Slovak
  56. Slovenian
  57. Spanish
  58. Swahili
  59. Swedish
  60. Tahitian
  61. Tamil
  62. Telugu
  63. Thai
  64. Tongan
  65. Turkish
  66. Ukrainian
  67. Urdu
  68. Vietnamese
  69. Welsh
  70. Yucatec Maya

    Community partners

Microsoft Translator has engaged with community partners to increase the number of languages and to improve overall language translation quality. Below is a list of community partners that Microsoft Translator has teamed with.
Additionally, Microsoft has teamed with the Klingon Language Institute, which promotes the constructed language, Klingon, which is used within the fictional Star Trek universe produced by Paramount and CBS Studios. Klingon has been supported by Microsoft Translator since May 2013.

SW development