Watch all the Transform 2020 sessions on-demand here.
Bound for a foreign country where you don’t speak the language? Good news: Google has you covered. This afternoon, the Mountain View tech giant announced that Translate, its free multilingual machine translation service, is now more robust. Offline, translation accuracy has improved by an order of magnitud in some cases.
Across the 59 languages supported offline by Translate on Android and iOS, accuracy is now 12% better, with enhanced word choice, grammar, and sentence structure. For some languages, including Japanese, Korean, Thai, Polish, and Hindi, the quality gain is more than 20%. In a related improvement, Translate now offers offline transliteration support for 10 new languages, including Arabic, Bengali, Gujrati, Kannada, Marathi, Tamil, Telugu and Urdu, such that now it provides an equivalent spelling in a familiar alphabet (for example, showing “नमस्ते” and “namaste” for “hello” where “namaste” is the transliteration of “नमस्ते.”)
The new and improved Translate is available as of today, Google says. Users who’ve used offline translation before will see a banner on the app’s home screen that will take them to the right place to update their offline files, but those who haven’t will have to head to the offline translation settings screen and tap the arrow next to the language name to download that language. (Each language package takes up about 35-45 MB, roughly equivalent in size to previous offline packages but at a higher quality.)
June 5th: The AI Audit in NYC
Join us next week in NYC to engage with top executive leaders, delving into strategies for auditing AI models to ensure fairness, optimal performance, and ethical compliance across diverse organizations. Secure your attendance for this exclusive invite-only event.
Translate’s online translations have been powered by neural machine translation (NMT) since 2016, and a few algorithms made their way to mobile platforms in 2018. Rather than the previous machine learning approach that provided interpretation by scanning phrases of a sentence, offline translations with NMT analyze entire chunks of text at once, allowing for more natural-sounding, grammatically sound, context-aware translations. Like the previously used phrase-based machine learning approach, NMT leverages hundreds of millions of example translations of things like articles, books, documents, and search results. But while there shouldn’t be a noticeable difference in quality from previous offline translations, Google notes that online translations will still be more accurate than offline translations, as the company made concessions to reduce the size of language packages.
The Translate improvements come after the rollout of Google Assistant’s interpreter mode to Android devices, which enables Assistant to recognize and speak up to two languages at once. (Interestingly, Google says the same technology that underpins Assistant’s bilingual capabilities powers interpreter mode.) Google’s conversational AI more recently became conversant in over 30 languages in 80 countries (up from 8 languages and 14 countries in 2017), and it added nine new AI-generated variants of English voice tailored to English speakers in the U.K., India, France, Germany, Japan, Netherlands, Norway, Korea, and Italy. Perhaps more significantly, a new and improved version of Assistant launched on the Pixel 4 in early fall. Its English model works offline and processes speech at “nearly zero” latency, delivering answers up to 10 times faster than on previous-generation devices.
