“We have been paying huge attention to India-first in the last few years to understand the key problems here,” said Ben Gomes, who leads the overall search for Google as vice-president. He is also a Google Fellow, a rare title given to those with exceptional expertise in their area of work within the company.
India is already one of the world’s largest Internet and mobile markets as out of its 1.3 billion population 400 million people already use the internet, while another 250 million are expected to join the World Wide Web by 2020. To make most of them, if not all, use Google, Gomes wants to figure out, how to let people search on Google in languages other than English and find answers by talking to Google with the device talking right back to them.
“We realise that we need to create answers to questions, not just throw up the ten blue links like in the past,” he said. “So, if you ask who the prime minister of India is, you want an answer to that and possibly get a conversation started.” It is the “beginnings” of the computer trying to communicate, answer follow-up questions and have a conversation, in the same way as the human brain does through some understanding of the language and having a broad model of the world, said Gomes who grew up in Bengaluru after being born in Tanzania. It’s not easy, but Google is at it. One can hear a conversation in a crowded bar because humans have a shared model of the world even though one doesn’t understand each word, Gomes said.
“These problems are not easy, we take them as easy because for human beings it is natural — our brains work that way,” he said. “For computers, it’s been almost impossible in the past, and now it’s becoming possible with advances like machine learning.”
Google is working on technologies such as machine learning to ensure real time language translation, especially in India, where already there are 230 million regional language users online with 170 million of them using messaging services. Gomes said that while keyboards in 27 Indian languages are already there, the next step is to get more content available in regional languages.
“We are beginning to get highquality translation using machine learning for lots of languages,” he said. “We want to specialise it for the things that people need in India such as queries on Bollywood actors, local cities, government officials, etc.” Gomes said that despite the advancement of Search over the years, there is still a long way to go. “We would like to understand complex language and answer more complex questions,” he said. “In all these languages, we would like to encourage more content creation so that people can find answers in their languages and use translation to get stuff from around the world.
We are not there just yet, but that’s where we would like to go.” Gomes agreed that the issue of fake news, which cropped up at the time of the US elections last year, is serious. “It is an issue that affects a small fraction — 0.2% — of the traffic, but it is an important kind of a query,” he said. Google has responded by changing its algorithms over 2,000 times in the last one year and deploying raters to ranks news sources based on how authoritative and relevant they were.