We want to be more intelligent in search

Written by Diksha Dutta | Diksha Dutta | Updated: Sep 3 2012, 08:45am hrs
It is hard to imagine Google search being simpler, easier and at the same time more exhaustive. However, yet again, the internet giant has come up with the concept of Knowledge Graph in India which will help in searching information even through your personal emails. Founder of the search concept, Amit Singhal, senior vice-president, Search and Google Fellow feels that search should be able to understand humanity and swim through personal as well as public information. In an interaction with Diksha Dutta, he discusses the significance and scope of Knowledge Graph in India. Excerpts:

Please explain to us the concept and essentials of Knowledge Graph

The Knowledge Graph is Googles growing model of hundreds of millions of people, places and things that make up our world. It is an ambitious technology that will help us realise our vision for the future of searchone where Google provides answers not just based on strings of letters and words, but also on an understanding of real-world things, their defining characteristics and their connections to one another.

Today, we have 500 million objects (people, places. things) in our Knowledge Graph, and know billions of facts about and connections between these objects. We are constantly growing and evolving the Knowledge Graph, and we want it to describe as much of the world as possible.

Why did Google start building a Knowledge Graph

When you come to search, you are not just looking for a webpage. You are looking to discover something you did not know about the world, or to get something done. In order to provide you with results that help you accomplish this, Google needs to understand not only keywords, but the meaning behind those words. When we can understand that those words refer to real-world things, we can do a better job of getting you the answers you need.

Why is this important to the future of search

To date search has been primarily about strings of words and matching strings to documents. We want it to be more intelligentto really understand the world of things, more like the way people do, so it can give you answers regardless of whether theres a webpage out there that already has the answer.

Is Knowledge Graph also incorporated in other geographies How has the success been

Currently the Knowledge Graph is available globally for queries in English. We want to extend the Knowledge Graph to as many regions and languages as we can around the world, over time. It's a really interesting technical challenge since it requires much more than just translating wordsthe Knowledge Graph needs to be tailored to the regionally specific meanings of things in the real world. So if you search for chiefs in the US, you get the Kansas City Chiefs, a football team in the US. But if you search for chiefs in Australia, you get the rugby team there. The same word, chiefs, refers to very different real-world entities.

Is this also an application for enterprises that Google will be offering separately How can this help enterprises

Google Search is a general product for all users.

Where does the information in Knowledge Graph come from

For the past couple years, we have been building a Knowledge Graph of hundreds of millions of real-world entities, learning from the collective wisdom of the Web, including sources like Freebase, Wikipedia, and the CIA Factbook. We also take into account what people on Google are most interested in learning about a particular topic. By looking at the most frequently asked questions about a topic, we can prioritise the facts and connections that we collect in the Knowledge Graph.

How does the Knowledge Graph change the way Google search engine interprets queries

One of our biggest challenges in building the perfect search engine is that computers do not understand the world like people do. A search engine sees strings of numbers and letters, while people see places, things, and other people. With Knowledge Graph we are taking a step away from the computer simply matching keywords toward teaching the computer to understand the world the way people do.

How can you ensure that the information in Knowledge Graph results is accurate

We try to reflect generally accepted attributes of and relationships between things, people and places. To do that, we typically draw from sources that are already public. Our goal is to be useful, we realise we will never be perfect, just as a persons librarys knowledge is never complete.