Much of the research that Google concerns itself with is in the field of artificial intelligence. From achieving a masterful level of proficiency in playing challenging board games to detecting medical anomalies in retinal scans to more recently being able to offer human-like interaction in chat apps, AI research is growing to be the engine .
Recently the search giant unveiled a new technique that will enable far more accurate language translation, something that is used across many of their products including Google Translate, Google Search, Google Goggles and in numerous instances of translation services embedded directly into their search results.
This approach, called the Google Neural Machine Translation system (GNMT), works on entire sentences as a whole as compared to earlier approaches that focused on phrases and individual words. With this new method, translations will be far more natural, as a native speaker of the language would expect them to be. The earlier word and phrase based translation systems often resulted in stilted results, or ones that conveyed an altogether incorrect intent.
Image courtesy: Google
Quoc V. Le and Mike Schuster, researchers on the Google Brain team, stated "The advantage of this approach is that it requires fewer engineering design choices than previous Phrase-Based translation systems."
This ‘macro’ approach also results in a direct savings on the amount of compute power utilized during such translations, which translates into better battery life on mobile devices in which they are run.
As with Neural Machine Translation systems, their accuracy increases with the amount of useful data they are fed with. In the case of GNMT, translation efficiency matched earlier systems at the outset, then proceeded to surpass them over time when tested with languages that have a near parity (eg. English to French or German.) The next challenge is to apply this new technique to more challenging languages such as Mandarin or Japanese.
Google Translate performs around 18 million translation per day across their mobile and Web apps.
){kind=small}
){kind=small}
){kind=small}
){kind=small}
){kind=small}
){kind=small}
){kind=small}
){kind=small}
){kind=small}
){kind=small}
){kind=small}
){kind=small}
){kind=small}
){kind=small}
){kind=small}
){kind=small}
){kind=small}
){kind=small}
 "Google Translate")
){kind=spacer}
){kind=small}
){kind=small}
){kind=small}
){kind=small}
){kind=small}
)
){kind=small}
){kind=small}
)
)
)
)
)
){kind=small}
){kind=small}
){kind=small}
){kind=small}
){kind=small}