Via 163.com [CN] (Google machine translation).
Via 163.com [CN] (Google machine translation) Update: We now have better pictures from 163.com (see below).
Last fall, Google introduced a new system for machine-assisted language translations, Google Neural Machine Translation system (GNMT), which takes advantage of deep neural networks to translate entire sentences – not just phrases – for greatly improved translations.
The Google company AI team recently announced they developed Google Neural Machine Translation system, GNMT, using a new technique that is improving results to near human translation speed accuracy.
Someday Google hopes to be able to translate both sides of a phone conversation as it happens, said Franz Och, head of Google's machine translation group.
The new system, which uses deep machine learning to mimic the functioning of a human brain, is called the Google Neural Machine Translation system, or GNMT.
The Google Neural Machine Translation system, deployed today for Chinese-English queries, is a step up in complexity from existing methods.
What's new here is that Google uses machine translation techniques to find obvious mistakes (see headline) as well as more subtle issues.
Quite often, though, Google Translate-style machine translation is good enough and there is clearly a huge market for all kinds of translation services.
Douglas Adams, Hitch-Hikers Guide to the Galaxy Google now offers instant machine translation in Google Talk, a step on the road to the ultimate Babel Fish.
That's how Google does its machine translation.
