 "Google introduces machine-learning products for nine Indian languages")
Aiming to bring a billion people online and make the web more useful for them, Google India on Tuesday unveiled new products on advancement in machine learning for Indian languages.
Google also announced that the neural machine translation is now available for nine Indian languages -- Hindi, Bengali, Punjabi, Marathi, Gujarati, Tamil, Telugu, Malayalam and Kannada.
"Google wants to extend the internet for every Indian. We have identified gaps that bar Indians from accessing the internet. There are 400 million internet users in India and the number is expected to reach 600 million by 2020," Rajan Anandan, Vice President, India and SouthEast Asia, Google, told reporters here.
He added that 300 million Indians access the internet on smartphones. Anandan also spoke about Google's tie-up with RailTel to provide high-speed internet at Railway stations.
The neural machine translation is available in Chrome and Maps to make the translation process seamless and refined.
Also Read
The company said it does one billion translations every day and 95 per cent of Google Translate has its usage outside of the US.
"Of over 500 million people who use Google Translate, most of the users are in India, Indonesia, Brazil and Thailand," it said.
According to Google, neural networks initially took 10 seconds to translate but the company worked on it and brought down the time to 0.2 seconds in two months.
The company also launched "Gboard" in 22 scheduled languages in India. The users can now search words, meanings and even emojis in local language.
The keyboard now has a new feature by which text editing can be done on the go easily.
The new Gboard comes with a feature that makes it easy to resize and reposition the keyboard according to a user's need.
Google also unveiled Hindi dictionary in Google Search in collaboration with Oxford University Press.
It also shared findings from a new report by Google and KPMG India, titled "Indian Languages-Defining India's Internet".
