Google announced its Translate app is adding its Neural Machine Translation (GNMT) technology to three more languages, bring natural-sounding interpretation to Hindi, Russian and Vietnamese, according to Engadget.
GNMT represents a revolutionary approach to machine translation, eschewing phrase-based techniques that break up sentences into small parts and instead taking a wider contextual look at the overall text This approach yields translations that not only are more accurate, but also more elegant and sound like the words of a fluent speaker. In fact, translations conducted by GNMT sometimes have a poetic quality.
For example, when translating from Hindi to English, a sentence processed using Google Translate’s previous technology came out as, “Here we have discovered with open eyes is no end to the adventures.” With GNMT, the same passage translates to, “If we search with open eyes then there is no end to adventure.”
To attain such high-quality translations, GNMT went to the extraordinary measure of creating its own language that allowed it to identify commonalities between different human tongues. Beyond making natural-sounding sentences, the GNMT’s machine language is so powerful it can perform translations between two languages it wasn’t specifically trained to support.
Google began applying its neural technology to Mandarin Chinese last year. It then expanded support to English, French, German, Japanese, Korean, Portuguese and Turkish.
Google said it will expand GNMT other more languages during the next few weeks. Eventually, Google plans to use neural machine translation for all 103 languages supported by the Translate app.
Pratik Dhebri is senior director of product management at Veritone and is truly democratizing Artificial Intelligence (AI) by orchestrating best-of-breed cognitive engines. He works with AI developers and cognitive engine providers to architect and monetize algorithms and applications. Learn more about our platform and join the Veritone developer ecosystem today.