From CNET: Google Neural Machine Translation (GNMT), first announced back in September, is the new system powering Google Translate, the platform that invalidates all those years of French classes. It's built on a neural network, a kind of computer system modeled on the human brain that can learn from past actions to solve new problems without being specifically programmed to do so.
While Google's already boasted about improved translation accuracy thanks to machine learning, the new Google Translate can translate from one language to another even if those two languages weren't paired by the system previously.
A post on the Google Research blog on Tuesday explained how:
"Let's say we train a multilingual system with Japanese to English and Korean to English examples... [GNMT] shares its parameters to translate between these four different language pairs... It can generate reasonable Korean to Japanese translations, even though it has never been taught to do so... To the best of our knowledge, this is the first time this type of transfer learning has worked in Machine Translation."
View: Article @ Source Site
