Word Lens uses Google’s machine learning algorithms, using computer vision to distinguish between letters on an image
Google today said its Translate app will now support offline translations and instant visual translation or reading things like road signs and hoardings in Bengali, Gujarati, Kannada, Marathi, Tamil, Telugu, and Urdu. The app has also added support for conversation mode- which will allow users to have a bilingual conversation with someone, simply by talking to the Google Translate- in regional languages.
Offline support enables Indian language users to translate a word or sentence even when they are not connected to the Internet. All the user has to do is download a language pack to their phone and get translations at any time.
Google also announced its Word Lens feature, which will let local Indian language speakers understand English street signs in the city or decide what to order from a restaurant menu. The feature will let users instantly translate text using their camera, in languages other than English. The Translate app already lets the user use camera mode to snap a photo of English text and get a translation for it in different languages.
Word Lens uses Google’s machine learning algorithms, using computer vision to distinguish between letters on an image.
The way it works is, the Translate app recognizes characters in a picture, which is where Google’s deep learning or neural networks come in.
Google has been focusing on building capabilities for local languages in India, which is the next growth frontier for mass adoption of the Internet. A report by industry body Internet and Mobile Association of India last year found that the English language still accounts for 56 percent of the content on the worldwide Web, while Indian languages account for less than 0.1 percent.
The report had identified two main challenges to mass adoption of the Internet in local languages- generating content in local languages and popularising such content for broader adoption, and the availability of Indian scripts for generating digital content.
“The first barrier is infrastructure,” Barak Turovsky, Group Product Manager, Translate, told Moneycontrol over a webcast on Wednesday. “For people who do not speak English, the Internet is not a very friendly place,” he added.
However, Google realises that there is also a literacy barrier that excludes a large population from taking advantage of the true potential of the Internet, and has been working on improving its speech recognition technology and speech-to-text capabilities, even in regional languages. According to the Internet giant, voice search in Hindi has seen a 400 percent year-on-year growth. It launched search feature in local Indian languages last month.
“A much higher percentage of users are using speech in India,” said Turovsky.
In April this year, Google launched translations between English and nine Indian languages — Hindi, Bengali, Marathi, Tamil, Telugu, Gujarati, Punjabi, Malayalam and Kannada. Neural network based learning essentially teaches a machine to think the way the human mind does, and translation uses much more context and becomes much more accurate, translating full sentences at a time, instead of pieces of a sentence.
Google said the number of its Translate app users has been doubling over the past six months.
How it works:
# Download the specific language pack from the Translate app when connected to the internet.
# Open the Translate app, make the language choice on either side of the screen and then tap the download button.
# To translate between two languages offline, you’ll need to download both language packs.
# Open the Translate app and point your camera at an English sign or text# The translated text in your language will get overlaid on your screen, even if you don't have an Internet or data connection.