Both human input through its Translate Community and neural machine translation (NMT) underpin Google Translate. A basic understanding of neural networks and deep learning is necessary for understanding NMT. In a nutshell, NMT leverages the millions of phrases that humans have already translated to create links between words at the sentence level, such as gender and number agreement, and then "learns" how to translate future sentences with a similar structure. In September 2016, Google unveiled their own version of the technology, dubbed Google Neural Machine Translation (GNMT), which represented a major improvement over earlier approaches like statistically based machine translation.
The Google Translate Community operates similarly to Google's Local Guides, who post ratings and details on cafes, restaurants, and attractions on Google, including their hours of operation and costs. Volunteers who assess translations and provide quick translations of a few phrases make up the Translate Community. The translation engine has been fed millions upon millions of contributions, which ultimately aids in its ability to learn more. The GNMT is still a work in progress and does not pick up linguistic nuances audible solely to the human ear.
Komentar