Monday, March 12, 2018

Microsoft Translation AI: DNN vs. statistical

Machine translation of human languages has improved significantly in last 10 years, but it is still far from perfect. Here are very good articles explaining how this works now,
and interesting web site that compares various methods of translation for select languages.

Machine Translation - Microsoft Translator

Both SMT (Statistical Machine Translation) and
Deep Neural Network) translation technologies
have two elements in common:
  • Both require large amounts of pre-human translated content (up to millions of translated sentences) to train the systems.
  • Neither act as bilingual dictionaries, translating words based on a list of potential translations, but translate based on the context of the word that is used in a sentence."
Try and Compare - Microsoft Translator
Artificial Intelligence, powered by neural networks:
used to be about 20 languages, now only Chinese; site sometimes works

Classic "statistical" translator: about 60 languages; site works, and this is same as API

marketing pages:

Try Microsoft Translator for free:
Bing, Live conversations, Office 365 Word,

Translator Text API - Auto Translate | Microsoft Azure

API documentation:

sample code calling Microsoft Translation API (classic):

Google paper: