In basic terms, the new Neural Machine Translation efforts allow Google Translate to take on whole sentences at one time, rather than the previous method which translated each word one at a time. This new system, according to Google, allows for machine learning to create a translated sentence that contains better grammar versus the old method, and it will continue to learn how to make better sentences over time.
With today’s update, Google Translate will use the new Neural Machine Translation methods on text to and from English in French, German, Spanish, Portuguese, Chinese, Japanese, Korean and Turkish. The company claims that this is the single biggest improvement for Google Translate in the last 10 years. The ultimate goal is for the Neural Machine Translation method to be extended to the over 100 languages that are currently supported by Google Translate.
If you use Google Translate, what are your impressions of the translations in the languages that are supported by the new machine learning method?