10 years later, in 2016, the company developed its own framework, Google neural machine translation technology (GNMT). Google knew they needed to swap machine translation (MT) technologies to improve its accuracy. But it began to offer odd translations for longer, complex sentences. Since it translated individual words, it worked most efficiently for brief phrases. But, it soon became clear that this wasn’t going to work in the long term. When it first launched in 2006, it used statistical machine translation to provide instant translated text. Apart from that, it relies on various digital resources and common translations for languages. Ever wondered how Google Translate built its translation database? It’s heavily based on the Europarl Corpus, which is a collection of documents from European Parliament procedures that humans translated.
0 kommentar(er)
