Everything You Should Know About Google’s Neural Machine Translations http://blog.konkanitube.in/wp-content/uploads/2017/03/everything-you-should-know-about-googles-neural-machine-translations.gif
Google added Neural Machine Translation (NMT) to its Translate App last year in its ‘Landmark update’ for 8 different languages, and starting today, Hindi language users will be able to reap advantages that the new technology entails.
What’s Neural Machine Translation (NMT) feature in Google Translate?
In simple terms, Google Neural Machine Translation facilitates reading and translating entire sentence at one rather than translating individual words and phrases. This broader sentence context helps it figure out how the words actually have to be arranged in the translation, thus resulting translations are closer than ever to the way humans interact.
So, how effective is Neural Machine Translation? Well, Google says that this could result in 55 to 85 percent improvement over Statistical machine translation (SMT) that it had been using in Google Translate over the last 10 years! The picture below explains it better.
This is made possible by the vast amount of effort that Google has put in devising deep machine learning and AI based end-to-end systems that learn with time to improve their accuracy.
Why is Neural Machine Translation Important for Google?
Google wants to be present on every connected device in your living room and wants the software on these gadgets to be smart enough to interact with you just like you would with any human assistant, i.e. by using Voice Commands.
Therefore, it is of utmost importance it’s AI systems train themselves to understand intricacies of different languages and learn to respond pertinently. Also, this expertise in natural language understanding will make every Google service – Including Google Assistant, Google Translate, and even Google Search – better.
What languages support Google Neural Machine Translations?
Google started off with 8 languages: French, German, Spanish, Portuguese, Chinese, Japanese, Korean and Turkish. Now, Google has added Hindi, Vietnamese and Russian to this list.
A few other languages will be added in the next couple of weeks. Of course, eventually, Google plans to scale Neural Machine Translations to all 103 languages that it deals with.
Apart From Google, who else is using NMT?
Google isn’t alone in these endeavors. Recently, an App called ‘Papago’ from Navar Labs is Korean made app that relies on Neural Machine Translation and is currently much more accurate than Google at translating from English to South Korean and Simplified Chinese.