Mandarin Chinese is one of the most difficult languages to learn. While online translation services can be a useful crutch, everyone is familiar with the frustrating inaccuracies these often produce.
Google hopes to consign this problem to the past as it launches ‘Google Neural Machine Translation’ (GMNT) for its popular product, Google Translate.
The tech giant’s previous translation system used a rudimentary ‘Phrase-Based Machine Translation.' This system worked by breaking sentences into individual words and phrases, converting them independently and thus rendering translations broken and disjointed when read as a whole.
GMNT is producing truer translations by considering sentences in their entirety. The result is a 55-58 percent reduction in translation errors in certain language pairings. Google Translate web and mobile apps now use GMNT for the 18 million Chinese-English translations they process a day.
Here you can see how PBMT, GMNT and human translation stack up against each other.
Clearly the difference is enormous, but Google have acknowledged that the service is still some way off perfect and are working on ways of further increasing its accuracy by placing sentences in the context of paragraphs and even whole pages.
0 User Comments