Microsoft translator should be integrated as match resource into translate5 analogous to Moses, SDL languageCloud and Google translator.

The API documentation can be found here:

https://docs.microsoft.com/en-us/azure/cognitive-services/translator/

Usage of dicitionary API for smaller phrases

MS Translator is the only MT engine, that is able to send along potential synonyms / alternative translations and back translations, when the dictionary API of MS is used.

Therefore if a phrase is max 30 characters, instead of the normal translation API of Microsoft translate5 will use the dictionary API of Microsoft. If the dictionary API does return an empty result, the normal translation api is queried.

To usual REST-calls coming from the translate5 GUI the answer of the MS match resource will look the same as if the normal translation API was queried (or here still the normal translation API is queried - what ever is more easy to implemenet).

Yet for the InstantTranslate integration also the following information is delivered to the GUI and displayed there:

• alternative translations
  • back translations to the source language for every alternative translation, so the user can better understand, what the actual meaning of the alternative translation is
• part of speech
• the confidence of the alternative translation

The attached screen shows, how the extended InstantTranslate interface for translations coming from MS Translator should look like.

Document the microsoft translator in the confluence.