Say What? Neural Networks Give Google Translate a Boost

...

Google Translate often serves up some laughable results, but that could become a thing of the past with Google Neural Machine Translation (GNMT), which is currently at work on Chinese-to-English translations.

"Rapid advances in machine intelligence have improved our speech recognition and image recognition capabilities, but improving machine translation remains a challenging goal," Quoc V. Le and Mike Schuster, Google Brain Team research scientists, wrote in a blog post. GNMT "utilizes state-of-the-art training techniques to achieve the largest improvements to date for machine translation quality."

Google Neural Machine Translation (GNMT)

When Google Translate launched a decade ago, it relied on Phrase-Based Machine Translation. The algorithm simply looked up individual words, replaced them with the equivalent word in another language, and pieced them together for what were often incoherent sentences.

Neural Machine Translation, however, considers the entire input sentence as a unit—like you would comprehend a whole image rather than its individual pixels—taking into account the nuances of speech and meaning.

"The advantage of this approach is that it requires fewer engineering design choices than previous Phrase-Based translation systems," Le and Schuster wrote.

Google highlighted a Chinese-to-English translation produced by GNMT (above), which is just one of the more than 10,000 language pairs supported by Google Translate. "We will be working to roll out GNMT to many more of these over the coming months," Google said.

"Machine translation is by no means solved," the company concluded. GNMT can still make "significant errors" a human translator would never make, like dropping words, or mistranslating proper names and rare terms. "However, GNMT represents a significant milestone."

Categories
GAMES
0 Comment

Leave a Reply

Captcha image


RELATED BY

  • 5300c769af79e

    Weekly Rewind: Napping on planes, Lucid’s Air EV, Facebook and Obama tackle AI

    Fortunately, Zuck’s thespian skills (or lack thereof) are played for laughs in the video, which comes across as parody of a consumer tech device ad.The World Economic Forum figures more than seven million jobs could be lost to automation by 2020, with around two million gained.
  • 5300c769af79e

    Unified Security Management (USM) vs. SIEM: a Technical Comparison

    Download Get a full overview of the changing security landscape, and more importantly insight into the rapidly changing SIEM category, and the reasons that have led to those changes.To offer a complete picture of the changes to SIEM technology, it is valuable for some to understand the context of SIEM market and how (and why) AlienVault differentiates itself from the traditional approach.
  • 5300c769af79e

    OnePlus Will Let You Review the OnePlus 3 Early

    The program is called “The Lab,” and it works as a “peer review” program that involves 30 of OnePlus’ biggest fans reviewing the OnePlus 3 to deliver a “comprehensive community review.It’s the same story all small companies pitch to their biggest fans, which is fine.
  • 5300c769af79e

    The Show: Episode 110 - Damn, Google I/O

    Since we ran out of time last week, thanks to travel schedules and all that, it’s time to recap Google I/O 2016 on the DL Show.After a weekend off, we’ve had time to fully digest all that Google announced, which should mean more well-rounded thoughts, or something, right?