It applies to the following text translations:

English <-> Chinese Simplified, Traditional, and Cantonese

English <-> Japanese

Chinese (all) <-> Japanese

In addition to this new capability, all speech translations are already using neural translations for the ten supported speech translation languages (Arabic, Chinese (Mandarin), English, French, German, Italian, Japanese, Portuguese, Russian and Spanish) in both the Microsoft Translator live and Skype Translator features.

Neural translations generate more fluent and more human sounding translations by using the context of a full sentence rather than just a few adjacent words. Using a multi-dimensional representation (a) of each word for a given translation pair (e.g. English-Chinese), multiple layers of the neural network build a larger multi-dimensional representation of this word within the context of the full sentence (b). The last layer uses these representations to translate words with the most appropriate translation given the sentence and in the most logical order in the target language sentence(c).

Microsoft’s AI-powered translation system delivers the most significant improvement in machine translation quality since statistical machine translation became the industry standard ten years ago. These improvements in quality and fluency have resulted in translations that are the closest they have ever been to human-generated ones.  You can compare the improved translations on all 11 available languages at http://translate.ai.

The post Microsoft Translator Apps Switch all Chinese and Japanese Language translations to Neural Network Technology appeared first on Microsoft Translator Blog.