YOU immediately notice when chatting with Google Fellow Amit Singhal that he is a person who is passionate about what he does. His eyes beamed with excitement and pride when he talked about Google and its search technologies.
After 20 years in the industry since graduating from university — with half that time spent at Google — you can’t help but wonder what keeps him so enthusiastic.
“I love search (technology) plain and simple,” he said, a fact that’s recorded in a blog. “Even though I’ve been in the industry for so long, I still come to work every morning like a kid going to a candy store.”
In the beginning it was exciting for Amit to watch the science fiction of search research become science fact over the decades. Even more exciting for him is to watch the technologies developing further at Google.
“Who would have thought that searching the Web would move beyond text — beyond languages, even?” he said. Or that a search engine could be made to recognise its users and understand what they’re looking for?
“The semantic systems we are building is something I never expect to work on in my lifetime. But here we are,” he said.
Amit finds it hard to remember what the Internet was like in the days before YouTube videos and online news. Everything’s changing so fast, he said. “When we first started working on search, it was all about text. Now we’re cataloguing many kinds of media and mapping everything out.”
“We’re also breaching the language barrier,” he said. English is the most common language among Internet users, but there are many non-English speakers who are using the Web.
Learning a new language is difficult for a human, let alone a computer but Google has been able to improve local search over the years thanks to specific adjustments or features built to cater to individual language needs. Going forward, Google doesn’t just want to translate between, say ... English and Japanese. It wants to translate between Japanese and Chinese, Japanese and Urdu, Japanese and Swahili, etc.
“You get the picture? We want to translate between every possible combination of languages,” Amit said.
He said that to this end, Google has taken a different route. Instead of designing a system to understand language rules, Google has taken a statistical approach to translation.
“That way, when there are exceptions to the rule, our system can handle those as well,” Amit said.
“We feed our translation engine thousands of professionally translated documents to build thousands of correlations so that we can start predicting the best translation.”
Today’s search engine tries to anticipate your needs and interests, according to him.
By relying on the IP address information, for example, Google can provide more specific answers to users. For instance, when you type “pizza in Kyoto,” into the serach box, you will get a list of pizza restaurants in Kyoto and not San Francisco or anywhere else.
“We also use Web History to help our users personalise their searches based on their interests, as well as look at social networking connections to try to accurately predict what content they might be most interested in,” said Amit.
At Google, he said, we are working to make our search engine able to understand everyone’s queries better than a human being could.
“We are trying to teach computers to understand how humans think. This is the holy grail of search,” he said. “This is why I feel like a kid in a candy store everytime I go to work.” — ZAM KARIM
Did you find this article insightful?