The ability in translate Google more and more close to the human ones – Stadio24.com

Thanks to the new system, Google Neural Machine Translation System (GNMT), Google Translate will take into account the whole paragraph and make the translations to be less approximate, reducing errors by up to 85% compared to the current translation system, Phrase-Based Machine Translation (PBMT), that takes into account the individual words or phrases, without worrying much on the context.

The new system Neural Machine Translation Google “bridges the gap between human and machine Translation” have said the recercatori, according to which the system is the highest level of development or knowledge attained so far by a technology translation, in particular for the translations from English to French and from English to German, where the errors have been reduced by an average of 60 percent. Of course, the work does not end here, the dream is of course to reach almost the perfection, who knows, maybe one day it really arrives with artificial intelligence. the Google has faced the problem enabling millions of users around the world can use your application for translations snapshots on the Web and not only. The translation word-for-word, leads to results that would be preposterous. Experts in the field artificial intelligence who have read the technical article of 23 pages say that those of Google have not invented anything but they have done a great job, using all the “tricks” to the state of the art of the IA and through their work have shown that the approach to machine translation based on neural networks is much more efficient. “This method provides a good balance between the flexibility of the ‘characters’ delimited, and the efficiency of the ‘words’ are delimited, and manages a natural definition of rare words, and finally, improves the overall accuracy of the system.”

Mountain View are already experiencing the GNMT on the translation from chinese to English (one of the most complicated), but the researchers Big G is already at work to bring in the next few months, the system on the translation also about the translations of other languages.

Differently, the new system considers the period as a single part to be translated.