For us millennials, Google Translate has been there for our help in translating difficult phrases and even huge documents from some unknown languages. The translator can detect the language and translate it for us. It’s been a huge help for us. Some of us also learn pronunciations by listening to the voice text and simple phrases so that we can communicate. In a way, it kind of bridges the gap between various cultures. In September, Google Translate switched from  Phrase-Based Machine Translation(PBMT) to Google Neural Machine Translation (GNMT).

machine translation

What is Machine Translation?

It is usually referred to a system or an online portal that helps in language translations, like Google Translate and Microsoft Translator.

Let’s shift our focus to Microsoft’s Machine Translation.

Since the early 2010s, a new artificial intelligence technology, deep neural networks (aka deep learning), has allowed the technology of speech recognition to reach a quality level that allowed the Microsoft Translator team to combine speech recognition with its core text translation technology to launch a new speech translation technology.


Now the basic Microsoft Machine Translation had two phases

a) Statistical Machine Translation that was based on more than a decade of natural-language research at Microsoft. Rather than writing hand-crafted rules to translate between languages, modern translation systems approach translation as a problem of learning the transformation of text between languages from existing human translations and leveraging recent advances in applied statistics and machine learning.

b) Neural Network Translation are based on the foundation of deep learning and machine learning. It is AI driven. By leveraging the scale and power of Microsoft’s AI supercomputer, specifically the Microsoft Cognitive Toolkit, Microsoft Translator now offers neural network (LSTM) based translation that enables a new decade of translation quality improvement. These neural network models are available for all speech languages through the Microsoft Translator Speech API, on the try and compare site, and through the text API by using the ‘generalnn’ category ID.

Based on the neural-network training, each word is coded along a 500-dimensions vector (a) representing its unique characteristics within a particular language pair (e.g. English and Chinese). Based on the language pairs used for training, the neural network will self-define what these dimensions should be. They could encode simple concepts like gender (feminine, masculine, neutral), politeness level (slang, casual, written, formal, etc.), type of word (verb, noun, etc.), but also any other non-obvious characteristics as derived from the training data.


Microsoft Translator is just not limited to Text but it extends to Speech as well. This technology is exposed in the Translator live feature (, the Translator apps, Skype Translator and is also Initially made available only through the Skype Translator feature and in the Microsoft Translator apps on iOS and Android, this functionality is now available to developers with the latest version of thethrough an open REST-based API available on the Azure portal.

Although it may seem like a straightforward process at a first glance to build a speech translation technology from the existing technology bricks, it required much more work than simply plugging an existing “traditional” human-to-machine speech recognition engine to the existing text translation one.



For more information, visit

For all the latest trends, tech news, & reviews follow us on Instagram, TwitterFacebook

Also Read: Machine Learning for all