'Lost In Translation' No More As Google Announces Its Neural Machine Translation System
To overcome the shortcomings of traditional machine translation systems, Google’s research team has come with up with Google Neural Machine Translation system (GNMT), which utilizes state-of-the-art training techniques to achieve the largest improvements to date for machine translation quality. Complete research results are described in a#-Link-Snipped-#titled- Google’s Neural Machine Translation System: Bridging the Gap between Human and Machine Translation.
With the breakthrough, notoriously hard to deliver Chinese to English machine translations are made possible with the Google Translate mobile and web apps. End-to-end learning systems of GNMT can also overcome the weaknesses inherent in the phrase-based systems currently in use for machine translation.New system of deep learning model reduces translation errors by up to 87%.
The phrase-based translation system divides sentences into phrases, which it then translates individually. After this the whole series of phrases has to go through another layer of processing to ensure the correct word order where quality is still variable.
The new method however uses a total of 16 processors to first transform words into a value known as a vector. A vector roughly represents how related one word is to every other word in the vast dictionary of training materials (2.5 billion sentence pairs for English and French; 500 million for English and Chinese). The system uses vectors from the input language to come up with a string of possible translations which are ranked based on their probability of occurrence.
The tech giant is planning to gradually replace the system underlying all of its translation work—one language at a time. Though NMT have been employed in machine translations as early as 2010, it is the first time any company has clubbed and deployed all the advances together. Google’s translation tool is currently live for web and mobile queries and sure makes up for another effective milestone.
Source:#-Link-Snipped-#
With the breakthrough, notoriously hard to deliver Chinese to English machine translations are made possible with the Google Translate mobile and web apps. End-to-end learning systems of GNMT can also overcome the weaknesses inherent in the phrase-based systems currently in use for machine translation.New system of deep learning model reduces translation errors by up to 87%.
The phrase-based translation system divides sentences into phrases, which it then translates individually. After this the whole series of phrases has to go through another layer of processing to ensure the correct word order where quality is still variable.
The new method however uses a total of 16 processors to first transform words into a value known as a vector. A vector roughly represents how related one word is to every other word in the vast dictionary of training materials (2.5 billion sentence pairs for English and French; 500 million for English and Chinese). The system uses vectors from the input language to come up with a string of possible translations which are ranked based on their probability of occurrence.
The tech giant is planning to gradually replace the system underlying all of its translation work—one language at a time. Though NMT have been employed in machine translations as early as 2010, it is the first time any company has clubbed and deployed all the advances together. Google’s translation tool is currently live for web and mobile queries and sure makes up for another effective milestone.
Source:#-Link-Snipped-#
0